fix(agent-kit): separate session cache per agent

Each agent now maintains its own session cache file instead of sharing
a single agent-sessions.json. This prevents session ID conflicts when
multiple agents operate on the same thread+role pair.

Changes:
- getCachePath() now takes agentName parameter
- getCachedSessionId/setCachedSessionId require agentName as first param
- Cache files named <agent>-sessions.json (e.g., hermes-sessions.json)
- Agent wrappers inject their agent name into cache calls
- Add comprehensive tests for session cache isolation
- Handle malformed JSON gracefully (treat as empty cache)

Fixes #461

.cursor/rules/sync-readme.mdc

@@ -0,0 +1,67 @@
+# Sync README
+
+When updating README.md files in this monorepo, follow these conventions.
+
+## Scope
+
+- Root `README.md` — project overview and navigation hub
+- Per-package `packages/*/README.md` — each package self-contained
+
+## Root README Structure
+
+The root README should have these sections in order:
+
+1. **Title and one-liner** — stateless workflow engine driven by single-step CLI
+2. **Overview** — 2-3 paragraphs explaining what it does and key concepts
+3. **Architecture** — dependency layer diagram (text-based)
+4. **Packages** — table with ALL packages from packages/ directory, columns: Package, Description, Type (cli/lib/agent/app)
+5. **Quick Start** — install, build, register workflow, start thread, run step
+6. **CLI Reference** — brief command list, detailed usage in cli-workflow README
+7. **Development** — bun install / build / check / test
+
+## Per-Package README Structure
+
+Each package README should have:
+
+1. **Title** — package name
+2. **One-line description** — matching package.json
+3. **Overview** — what it does, where it sits in the architecture, dependencies
+4. **Installation** — bun add (for libs) or "included as binary" (for cli/agents)
+5. **API** (lib packages) — all exports from src/index.ts with type signatures, grouped by category, minimal usage examples
+6. **CLI Usage** (cli/agent packages) — command reference with examples
+7. **Internal Structure** — brief src/ file organization
+8. **Configuration** (if applicable)
+
+## Execution Steps
+
+### Step 1: Gather current state
+For each package read:
+- package.json (name, version, description, dependencies, bin)
+- src/index.ts (public API exports)
+- Existing README.md (preserve hand-written content worth keeping)
+
+### Step 2: Update root README
+- Ensure ALL packages in packages/ directory are listed in the table
+- Update CLI command reference from uwf --help output
+- Keep Quick Start examples valid
+
+### Step 3: Write/update each package README
+- Follow the per-package structure
+- API section MUST match actual src/index.ts exports — never invent
+- For agent packages: document CLI binary name, how it is invoked
+- For lib packages: document exported types and functions
+- Internal structure: list actual files in src/
+
+### Step 4: Verify
+- All relative links work
+- Package names match package.json
+- No references to removed/renamed packages
+- bun run build still passes
+
+## Guidelines
+
+- Only document what src/index.ts actually exports
+- Root README summarizes, package READMEs go into detail
+- Verify CLI examples against actual commands
+- Preserve existing good prose when updating
+- English for all README content

.env.example

@@ -1,40 +0,0 @@
-# ──────────────────────────────────────────────
-# Workflow Engine — Environment Variables
-# ──────────────────────────────────────────────
-# Copy this file to .env and fill in the values.
-
-# ── Cursor Agent ──
-
-# CLI command to invoke the Cursor agent (required for develop workflow)
-WORKFLOW_CURSOR_COMMAND=
-
-# Model override for Cursor agent
-WORKFLOW_CURSOR_MODEL=
-
-# Timeout in milliseconds for Cursor agent operations
-WORKFLOW_CURSOR_TIMEOUT=
-
-# ── Hermes Agent (used by develop tester/committer + solve-issue) ──
-
-# CLI command to invoke the Hermes agent (absolute path required)
-WORKFLOW_HERMES_COMMAND=
-
-# Model override for Hermes agent
-WORKFLOW_HERMES_MODEL=
-
-# Timeout in milliseconds for Hermes agent operations
-WORKFLOW_HERMES_TIMEOUT=
-
-# ── Storage ──
-
-# Override the workflow storage root directory
-# Default: ~/.uncaged/workflow
-WORKFLOW_STORAGE_ROOT=
-
-# Gateway secret for the serve command
-WORKFLOW_DASHBOARD_SECRET=
-
-# ── Display ──
-
-# Set to any value to disable colored output
-# NO_COLOR=1

.gitignore

@@ -11,3 +11,5 @@ solve-issue-entry.ts
 packages/workflow-template-develop/develop.esm.js
 .DS_Store
 *.py
+.claude
+tmp

.plan/335-setup-validate.md

@@ -0,0 +1,83 @@
+# Test Spec: uwf setup model connectivity validation (#335)
+
+## Context
+
+File: `packages/cli-workflow/src/commands/setup.ts`
+Test file: `packages/cli-workflow/src/__tests__/setup-validate.test.ts`
+
+After `cmdSetup` writes config, it should send a test chat completion request to verify the configured model is reachable. If validation fails, warn the user (don't abort — config is already saved).
+
+## Implementation Notes
+
+- Add a `validateModel(baseUrl, apiKey, model)` function that sends a minimal chat completion request (`POST /chat/completions` with `messages: [{role:"user",content:"hi"}]`, `max_tokens: 1`)
+- Returns `Result<void, string>` — ok if 2xx response, error with reason string otherwise
+- Use `AbortSignal.timeout(15_000)` for the request
+- Both `cmdSetup` and `cmdSetupInteractive` should call it after saving config
+- `cmdSetup` returns validation result in its return object: `{ ...existing, validation: { ok: true } | { ok: false, error: string } }`
+- `cmdSetupInteractive` prints a warning to console if validation fails, success message if it passes
+- Use the project logger (`createLogger`) — no raw `console.log` except in interactive CLI output (per CLAUDE.md)
+
+## Test Cases (vitest)
+
+### 1. `validateModel` — success path
+- Mock `fetch` to return `{ status: 200, ok: true, json: () => ({}) }`
+- Call `validateModel(baseUrl, apiKey, model)`
+- Assert returns `{ ok: true, value: undefined }`
+- Assert fetch was called with correct URL (`${baseUrl}/chat/completions`), correct headers (`Authorization: Bearer ${apiKey}`), correct body (model, messages, max_tokens: 1)
+
+### 2. `validateModel` — HTTP error (401 unauthorized)
+- Mock `fetch` to return `{ status: 401, ok: false, statusText: "Unauthorized" }`
+- Call `validateModel(baseUrl, apiKey, model)`
+- Assert returns `{ ok: false, error: <string containing "401"> }`
+
+### 3. `validateModel` — HTTP error (404 model not found)
+- Mock `fetch` to return `{ status: 404, ok: false, statusText: "Not Found" }`
+- Assert returns `{ ok: false, error: <string containing "404"> }`
+
+### 4. `validateModel` — network timeout
+- Mock `fetch` to throw `DOMException` with name `AbortError`
+- Assert returns `{ ok: false, error: <string containing "timeout" or "unreachable"> }`
+
+### 5. `validateModel` — network error (DNS failure, connection refused)
+- Mock `fetch` to throw `TypeError("fetch failed")`
+- Assert returns `{ ok: false, error: <string mentioning connectivity> }`
+
+### 6. `cmdSetup` — includes validation result on success
+- Mock global `fetch` for `/chat/completions` to succeed
+- Call `cmdSetup({ provider, baseUrl, apiKey, model, storageRoot })`
+- Assert returned object has `validation: { ok: true, value: undefined }`
+- Assert config files are still written (existing behavior preserved)
+
+### 7. `cmdSetup` — includes validation result on failure (config still saved)
+- Mock global `fetch` for `/chat/completions` to return 401
+- Call `cmdSetup({ ... })`
+- Assert returned object has `validation: { ok: false, error: ... }`
+- Assert `config.yaml` and `.env` are still written (validation failure doesn't prevent saving)
+
+### 8. `cmdSetupInteractive` — prints success message on validation pass
+- Mock `fetch` for both `/models` and `/chat/completions` to succeed
+- Mock stdin to provide valid selections
+- Capture console output
+- Assert output contains a success message like "Model verified" or "✓"
+
+### 9. `cmdSetupInteractive` — prints warning on validation failure
+- Mock `fetch`: `/models` succeeds, `/chat/completions` returns 401
+- Mock stdin for valid selections
+- Capture console output
+- Assert output contains a warning about model not being reachable and suggests trying a different model
+
+### 10. `validateModel` — request body correctness
+- Mock `fetch` to capture the request body
+- Call `validateModel(baseUrl, apiKey, "test-model")`
+- Assert body is `{ model: "test-model", messages: [{role: "user", content: "hi"}], max_tokens: 1 }`
+
+## Export Requirements
+
+- `validateModel` must be exported (for direct unit testing)
+- Signature: `async function validateModel(baseUrl: string, apiKey: string, model: string): Promise<Result<void, string>>`
+- `Result` type: `{ ok: true; value: T } | { ok: false; error: E }` (project convention)
+
+## Files to Create/Modify
+
+- **New**: `packages/cli-workflow/src/__tests__/setup-validate.test.ts` — all test cases above
+- **Modify**: `packages/cli-workflow/src/commands/setup.ts` — add `validateModel`, integrate into `cmdSetup` and `cmdSetupInteractive`

.workflows/solve-issue.yaml

@@ -0,0 +1,197 @@
+name: "solve-issue"
+description: "TDD-driven issue resolution for small, focused changes. Loop protection relies on engine maxRounds."
+roles:
+  planner:
+    description: "Analyzes issue and outputs a TDD test spec"
+    goal: "You are a planning agent. You analyze Gitea issues and produce a TDD test specification that downstream roles will implement and verify."
+    capabilities:
+      - issue-analysis
+      - planning
+    procedure: |
+      On first run (no previous steps):
+      1. Read the issue and all comments from Gitea using `tea issues <number> -r <owner/repo>`
+      2. Read CLAUDE.md (or equivalent project conventions file) to understand coding standards
+      3. Assess whether the issue has enough information to produce a test spec
+      4. If insufficient info: comment on the issue via `echo "..." | tea comment <number> -r <owner/repo>` (skip if you already commented), then output status=insufficient_info and terminate
+      5. If sufficient: produce a detailed TDD test spec in markdown covering all scenarios
+
+      On subsequent runs (bounced back by tester with fix_spec):
+      1. Read the tester's output from the previous step to understand what's wrong with the spec
+      2. Revise the test spec accordingly
+
+      After producing the test spec:
+      1. Store it via `uwf cas put-text "<markdown content>"` and capture the returned hash
+      2. Put the hash in frontmatter.plan (required when status=ready)
+    output: "Output a brief summary of the test spec. Frontmatter must include: status (ready or insufficient_info) and plan (CAS hash of the test spec, required when status=ready)."
+    frontmatter:
+      type: object
+      properties:
+        status:
+          type: string
+          enum: [ready, insufficient_info]
+        plan:
+          type: string
+      required: [status]
+  developer:
+    description: "TDD implementation per test spec"
+    goal: "You are a developer agent. You implement code changes following TDD — write tests first, then implementation."
+    capabilities:
+      - coding
+    procedure: |
+      Before starting any work, ensure a clean worktree:
+      1. `git checkout main && git pull` to get the latest code
+      2. `git checkout -b fix/<issue-number>-<short-description>` to create a fresh branch
+         - If bounced back from reviewer or tester, reuse the existing branch and rebase onto latest main:
+           `git checkout main && git pull && git checkout <branch> && git rebase main`
+
+      Then implement TDD:
+      3. Read the test spec from CAS: `uwf cas get <plan hash>` (find the hash from the latest planner step's frontmatter.plan)
+      4. If bounced back from reviewer or tester: read the previous role's output to understand what needs fixing
+      5. Write tests first based on the spec
+      6. Implement the code to make tests pass
+      7. Ensure `bun run build` passes with no errors
+      8. Run `bun test` to verify all tests pass
+    output: "List all files changed and provide a summary. Frontmatter must include: status (done or failed)."
+    frontmatter:
+      type: object
+      properties:
+        status:
+          type: string
+          enum: [done, failed]
+      required: [status]
+  reviewer:
+    description: "Code standards compliance check"
+    goal: "You are a code reviewer. You verify code standards compliance — NOT functionality (that's the tester's job)."
+    capabilities:
+      - code-review
+      - static-analysis
+    procedure: |
+      Before reviewing, verify the git branch:
+      1. Run `git branch --show-current` — confirm the branch name references the issue number being worked on
+      2. If the branch doesn't correspond to the issue, flag it in your output and reject
+
+      Then perform code review:
+      Hard checks (must all pass):
+      3. `bun run build` — no build errors
+      4. `bunx biome check` — no lint violations
+      5. TypeScript strict mode — no type errors
+
+      Soft checks (review against CLAUDE.md conventions):
+      - Functional-first: `function` + `type`, not `class` + `interface`
+      - No optional properties (`?:`) — use `T | null`
+      - Naming conventions (kebab-case files, PascalCase types, camelCase functions)
+      - Module boundary discipline (folder exports via index.ts)
+      - No `console.log` (use structured logger)
+      - No dynamic imports in production code
+
+      Only review standards compliance. Do NOT test functionality.
+      If rejecting, you MUST explain the specific reason in your output.
+    output: "Explain your decision with specific file/line references. Frontmatter must include: approved (true or false)."
+    frontmatter:
+      type: object
+      properties:
+        approved:
+          type: boolean
+      required: [approved]
+  tester:
+    description: "Functional correctness verification"
+    goal: "You are a tester agent. You verify that the implementation correctly satisfies every scenario in the test spec."
+    capabilities:
+      - testing
+    procedure: |
+      1. Run `bun test` for automated test verification
+      2. Read the test spec from CAS: `uwf cas get <plan hash>` (find the hash from the latest planner step's frontmatter.plan)
+      3. Verify each scenario in the spec is covered and passing
+      4. Determine outcome:
+         - passed: all scenarios verified, tests pass
+         - fix_code: tests fail or implementation doesn't match spec → send back to developer
+         - fix_spec: the spec itself is wrong or incomplete → send back to planner
+    output: "Report test results per scenario. Frontmatter must include: status (passed, fix_code, or fix_spec)."
+    frontmatter:
+      type: object
+      properties:
+        status:
+          type: string
+          enum: [passed, fix_code, fix_spec]
+      required: [status]
+  committer:
+    description: "Commits and creates PR"
+    goal: "You are a committer agent. You create a clean commit and push a PR linking the original issue."
+    capabilities: []
+    procedure: |
+      Note: You inherit the developer's worktree and branch. Do NOT create a new branch.
+      1. Stage all changes: `git add -A`
+      2. Commit with a descriptive message referencing the issue: `git commit -m "type: description\n\nFixes #N"`
+      3. Push the branch: `git push -u origin <branch-name>`
+         - If push hook fails: capture the error log in your output, mark hook_failed
+      4. On push success: create a PR via `tea pr create --title "..." --description "..."`
+         - PR description must follow the project template: What / Why / Changes / Ref sections, with `Fixes #N` in Ref
+    output: "Include PR URL on success or error log on failure. Frontmatter must include: success (true or false)."
+    frontmatter:
+      type: object
+      properties:
+        success:
+          type: boolean
+      required: [success]
+conditions:
+  insufficientInfo:
+    description: "Planner determined there's not enough info to proceed"
+    expression: "$last('planner').status = 'insufficient_info'"
+  devFailed:
+    description: "Developer failed to implement"
+    expression: "$last('developer').status = 'failed'"
+  rejected:
+    description: "Reviewer rejected the implementation"
+    expression: "$last('reviewer').approved = false"
+  fixCode:
+    description: "Tester found code issues"
+    expression: "$last('tester').status = 'fix_code'"
+  fixSpec:
+    description: "Tester found spec issues"
+    expression: "$last('tester').status = 'fix_spec'"
+  hookFailed:
+    description: "Push hook failed"
+    expression: "$last('committer').success = false"
+graph:
+  $START:
+    - role: "planner"
+      condition: null
+      prompt: "Analyze the issue and produce an implementation plan."
+  planner:
+    - role: "$END"
+      condition: "insufficientInfo"
+      prompt: "Insufficient information to proceed; end the workflow."
+    - role: "developer"
+      condition: null
+      prompt: "Implement the plan from the planner."
+  developer:
+    - role: "$END"
+      condition: "devFailed"
+      prompt: "Development failed; end the workflow."
+    - role: "reviewer"
+      condition: null
+      prompt: "Send the implementation to the reviewer."
+  reviewer:
+    - role: "developer"
+      condition: "rejected"
+      prompt: "Reviewer rejected the implementation; fix the issues."
+    - role: "tester"
+      condition: null
+      prompt: "Review passed; run tests on the implementation."
+  tester:
+    - role: "developer"
+      condition: "fixCode"
+      prompt: "Tests found code issues; return to developer."
+    - role: "planner"
+      condition: "fixSpec"
+      prompt: "Tests found spec issues; return to planner."
+    - role: "committer"
+      condition: null
+      prompt: "Tests passed; commit and push the changes."
+  committer:
+    - role: "developer"
+      condition: "hookFailed"
+      prompt: "Push hook failed; return to developer to fix."
+    - role: "$END"
+      condition: null
+      prompt: "Commit succeeded; complete the workflow."

CLAUDE.md

@@ -2,46 +2,41 @@
 
 ## Project Overview
 
-This monorepo implements a workflow engine that executes single-file ESM bundles. Each workflow is a self-contained `.esm.js` file with an XXH64 hash as its version identifier. Shared types live in `@uncaged/workflow-protocol`; bundle authors typically depend on `@uncaged/workflow-runtime`.
+This monorepo implements a stateless workflow engine driven by a single-step CLI (`uwf`). Workflows are **YAML definitions** stored as CAS nodes; threads are immutable chains of CAS-linked step nodes. No daemon — each `uwf thread step` invocation runs one moderator→agent→extract cycle and exits.
 
 ### Key Terms
 
 | Concept | What it is |
 |---------|-----------|
-| **Workflow** | A single-file ESM module that exports `run` (workflow function) and `descriptor` (metadata). Identified by its XXH64 hash (Crockford Base32). |
-| **Bundle** | The physical `.esm.js` file stored in `~/.uncaged/workflow/bundles/`. |
-| **Thread** | A single execution of a workflow, identified by a ULID. State lives in CAS (linked nodes); active threads indexed in `threads.json`; completed rows in `history/*.jsonl`. Debug logs use `.info.jsonl`. |
-| **Role** | A named actor within a workflow. Each role produces output with typed `meta`. |
-| **Registry** | `workflow.yaml` — maps workflow names to current/historical bundle hashes. |
+| **Workflow** | A YAML definition (`WorkflowPayload`) with roles, conditions, and a routing graph. Stored as a CAS node, identified by its XXH64 hash. |
+| **Thread** | A single execution of a workflow, identified by a ULID. State is an immutable CAS chain; active threads indexed in `threads.yaml`; completed threads in `history.jsonl`. |
+| **Role** | A named actor within a workflow. Each role has a system prompt and a JSON Schema `outputSchema`. |
+| **Moderator** | JSONata-based graph evaluator — determines the next role (or `$END`) with zero LLM cost. |
+| **Agent** | An external CLI command (`uwf-hermes`, etc.) spawned by `uwf thread step`. Produces frontmatter markdown output. |
+| **CAS** | Content-Addressed Storage via `@uncaged/json-cas` — all workflow definitions, thread nodes, and outputs are immutable CAS nodes. |
+| **Registry** | `~/.uncaged/workflow/registry.yaml` — maps workflow names to current CAS hashes. |
 
 ### Monorepo Structure
 
 ```
 workflow/
   packages/
-    workflow-protocol/              # @uncaged/workflow-protocol — shared types + Result
-    workflow-runtime/               # @uncaged/workflow-runtime — createWorkflow, type re-exports
-    workflow-util/                  # @uncaged/workflow-util — Base32, ULID, logger, storage paths, refs helpers
-    workflow-reactor/               # @uncaged/workflow-reactor — LLM fn + thread reactor (tool calls)
-    workflow-cas/                   # @uncaged/workflow-cas — CAS store, hash, Merkle
-    workflow-register/              # @uncaged/workflow-register — bundle validation, registry YAML, model resolution
-    workflow-execute/               # @uncaged/workflow-execute — engine, extract, fork, GC, workflowAsAgent
-    cli-workflow/                   # @uncaged/cli-workflow — uncaged-workflow CLI
-    workflow-agent-cursor/          # @uncaged/workflow-agent-cursor
-    workflow-agent-hermes/          # @uncaged/workflow-agent-hermes
-    workflow-agent-llm/             # @uncaged/workflow-agent-llm
-    workflow-agent-react/             # @uncaged/workflow-agent-react
-    workflow-util-agent/            # @uncaged/workflow-util-agent — buildAgentPrompt, spawnCli
-    workflow-template-develop/      # @uncaged/workflow-template-develop
-    workflow-template-solve-issue/  # @uncaged/workflow-template-solve-issue
-    workflow-dashboard/             # @uncaged/workflow-dashboard — React dashboard (private app)
-  docs/             # RFCs, conventions
-  biome.json        # root Biome config
-  tsconfig.json     # root TypeScript config
+    workflow-protocol/    # @uncaged/workflow-protocol — shared types (WorkflowPayload, StepNodePayload, WorkflowConfig, etc.)
+    workflow-util/        # @uncaged/workflow-util — Crockford Base32, ULID, logger, frontmatter parsing/validation
+    workflow-moderator/   # @uncaged/workflow-moderator — JSONata graph evaluator
+    workflow-agent-kit/   # @uncaged/workflow-agent-kit — createAgent factory, context builder, extract pipeline
+    workflow-agent-hermes/ # @uncaged/workflow-agent-hermes — uwf-hermes CLI binary (spawns hermes chat)
+    cli-workflow/         # @uncaged/cli-workflow — uwf CLI binary
+  legacy-packages/       # Archived packages (preserved for reference, not active)
+  examples/              # Workflow YAML examples (solve-issue.yaml)
+  docs/                  # Architecture docs
+  biome.json             # root Biome config
+  tsconfig.json          # root TypeScript config
 ```
 
-- Execution stack layers: `workflow-protocol` → (`workflow-runtime`, `workflow-util`, `workflow-reactor`) → (`workflow-cas`, `workflow-register`) → `workflow-execute` → `cli-workflow`
+- Dependency layers: `workflow-protocol` → (`workflow-util`, `workflow-moderator`) → `workflow-agent-kit` → `workflow-agent-hermes` / `cli-workflow`
 - Packages use `workspace:^` protocol (resolves to `^x.y.z` on publish)
+- External CAS: `@uncaged/json-cas` (store API, hashing, schema validation) + `@uncaged/json-cas-fs` (filesystem backend)
 
 ## Language & Paradigm
 
@@ -109,8 +104,6 @@ type WorkflowEntry = {
 - Always named exports, never default exports
 - One module = one responsibility, filename = purpose
 
-Workflow bundles (`.esm.js`) follow the same rule: export `const run` and `const descriptor`, not `export default`.
-
 ### Folder Module Discipline
 
 Every folder under `src/` is a **module boundary**. Four rules:
@@ -136,10 +129,10 @@ export { createCasStore } from "../cas/cas.js";
 
 // ❌ Bad — types defined in index.ts
 // in cas/index.ts:
-export type CasStore = { ... };  // should be in cas/types.ts
+export type CasStore = { ... }; // should be in cas/types.ts
 ```
 
-**Exception**: The package-level `src/index.ts` is the public API surface and re-exports from folder `index.ts` files. Files that remain at `src/` root (e.g. `types.ts`, `workflow-as-agent.ts`) are not inside a folder module and follow normal rules.
+**Exception**: The package-level `src/index.ts` is the public API surface and re-exports from folder `index.ts` files. Files that remain at `src/` root (e.g. `types.ts`) are not inside a folder module and follow normal rules.
 
 ## Naming
 
@@ -160,7 +153,7 @@ Workflow names use **verb-first** kebab-case:
 ### ID Encoding
 
 All IDs use **Crockford Base32**:
-- Bundle hash: XXH64 → 13-char Crockford Base32
+- CAS hash: XXH64 → 13-char Crockford Base32
 - Thread ID: ULID → 26-char Crockford Base32 (10 timestamp + 16 random)
 
 ## Error Handling
@@ -189,7 +182,7 @@ import { createLogger } from "@uncaged/workflow-util";
 const log = createLogger();
 
 // Each call site has a fixed 8-char Crockford Base32 tag
-log("4KNMR2PX", "Loading workflow bundle...");
+log("4KNMR2PX", "Loading workflow...");
 log("7BQST3VW", `Role ${role} started`);
 ```
 
@@ -204,7 +197,7 @@ log("7BQST3VW", `Role ${role} started`);
 
 ### Why fixed tags?
 
-- `grep "4KNMR2PX"` in `.info.jsonl` → instant code location
+- `grep "4KNMR2PX"` in logs → instant code location
 - No need for file/line info in the log — tag is the locator
 - Survives refactoring (tag stays the same when code moves)
 
@@ -221,74 +214,76 @@ console.log(result);
 
 Do NOT use `await import()` in production code. Always use static top-level `import`.
 
-**Exception**: The bundle loader and `extractBundleExports` dynamically import user workflow files at runtime.
-
-```ts
-// Dynamic import required: user bundle path resolved at runtime
-const mod = await import(bundlePath);
-```
-
 Test files (`__tests__/**`) are exempt.
 
 ## Toolchain
 
 | Tool | Purpose |
 |------|---------|
-| **bun** | Package manager + runtime + test runner |
+| **bun** | Package manager + runtime |
 | **TypeScript** | Type checking (strict mode) |
 | **Biome** | Lint + format (replaces ESLint + Prettier) |
+| **vitest** | Test runner (`cli-workflow` uses vitest; other packages use `bun test`) |
 
-### Commands
+### Development Workflow
 
 ```bash
-bun run check       # tsc --build + biome check
-bun run format      # biome format --write
-bun test            # run tests
+# ── Setup ──
+bun install                 # install all workspace dependencies
+
+# ── Daily development ──
+bun run build               # tsc --build (all packages, dependency order)
+bun run check               # tsc --build + biome check + lint-log-tags
+bun run format              # biome format --write
+bun test                    # run tests across all packages
+
+# ── Before committing ──
+bun run check               # must pass — typecheck + lint + log tag validation
+bun test                    # must pass — all package tests
 ```
 
-### Version Management & Publishing
+### Publishing
 
-All public `@uncaged/*` packages are published to **npmjs.org** via `@changesets/cli` with **fixed mode** (all packages share the same version number). `workflow-dashboard` is private and excluded.
+All public `@uncaged/*` packages are published to **npmjs.org** with **fixed mode** (all packages share the same version number).
 
 ```bash
-# 1. After making changes, add a changeset describing the change
+# 1. Add a changeset describing the change
 bun changeset
 
-# 2. Before release, bump all package versions + generate CHANGELOGs
+# 2. Bump all package versions + generate CHANGELOGs
 bun version
 
-# 3. Build, test, and publish to npmjs
+# 3. Build, test, and publish (runs scripts/publish-all.mjs)
 bun release
+
+# Or publish manually with a tag:
+node scripts/publish-all.mjs --tag alpha
+node scripts/publish-all.mjs --dry-run    # preview without publishing
 ```
 
 - `workspace:^` dependencies resolve to `^x.y.z` on publish
+- Publish order defined in `scripts/publish-all.mjs` (dependency order)
 - Changesets config: `.changeset/config.json` (fixed mode, public access)
-- Each package has auto-generated `CHANGELOG.md`
 
-### Consuming @uncaged/* Packages
-
-External workflow repos just `bun install` — packages come from npmjs like any other dependency. No special registry config needed.
-
-### End-to-end: Monorepo → Registry → Workspace → Bundle
+### End-to-end: Author → Register → Run
 
 ```
-workflow/ (monorepo)           — engine, runtime, templates, agents
-  │  bun release               — build + test + changeset publish
+examples/solve-issue.yaml       — write a workflow YAML definition
+  │  uwf workflow put
   ▼
-npmjs.org                      — @uncaged/* scoped packages (public)
-  │  bun install
+~/.uncaged/workflow/cas/        — Workflow stored as CAS node
+~/.uncaged/workflow/registry.yaml — name → hash mapping updated
+  │  uwf thread start <name> -p "..."
   ▼
-my-workflows/ (workspace)     — normal package.json
-  │  bun run build:develop     — bun build → single .esm.js
+~/.uncaged/workflow/threads.yaml — new thread head pointer
+  │  uwf thread step <thread-id>
   ▼
-uncaged-workflow workflow add  — register bundle locally
-uncaged-workflow run           — execute workflow
+moderator → agent → extract      — one step per invocation, repeat until $END
 ```
 
-1. **Monorepo changes** → `bun changeset` (describe change) → `bun version` (bump) → `bun release` (publish)
-2. **Workspace** → `bun install` fetches latest from npmjs
-3. **Build** → produces single-file ESM bundle with `@uncaged/*` as externals
-4. **Register & Run** → `uncaged-workflow workflow add <name> <bundle>` then `uncaged-workflow run <name>`
+1. **Author** — write a workflow YAML file with roles, conditions, and graph
+2. **Register** — `uwf workflow put <file.yaml>` parses YAML, registers output schemas, stores `WorkflowPayload` in CAS
+3. **Run** — `uwf thread start` creates a thread, `uwf thread step` executes one cycle per invocation
 
 ## Commit Convention
 
@@ -296,5 +291,5 @@ uncaged-workflow run           — execute workflow
 <type>(<scope>): <description>
 
 type: feat | fix | refactor | docs | chore | test
-scope: workflow | cli | rfc-001 | ...
+scope: workflow | cli | moderator | agent-kit | hermes | util | protocol | ...
 ```

README.md

@@ -1,71 +1,103 @@
 # @uncaged/workflow
 
-A workflow engine that executes single-file ESM bundles. Each workflow is a self-contained `.esm.js` file identified by its XXH64 hash (Crockford Base32).
+A stateless workflow engine driven by a single-step CLI. Workflows are YAML definitions with roles, JSONata routing conditions, and a directed graph. Threads are immutable CAS-linked chains — each `uwf thread step` runs one moderator→agent→extract cycle and exits.
 
-## Core Concepts
+## Overview
 
-| Concept | Description |
-|---------|-------------|
-| **Workflow** | A single-file ESM module exporting `run` (workflow function) and `descriptor` (metadata). Identified by its XXH64 hash. |
-| **Bundle** | The physical `.esm.js` file stored in `~/.uncaged/workflow/bundles/`. |
-| **Thread** | A single execution of a workflow, identified by a ULID. CAS-backed chain plus `threads.json` / `history/*.jsonl`; `.info.jsonl` for debug logs. |
-| **Role** | A named actor within a workflow. Each role produces output with typed `meta`. Roles live inside template packages (`src/roles/`). |
-| **Registry** | `workflow.yaml` — maps workflow names to current/historical bundle hashes. |
-| **CAS** | Content-Addressed Storage — bundles are immutable and addressed by hash. |
+This monorepo implements **uwf**, a workflow engine with no long-running daemon. You register YAML workflow definitions in a content-addressed store (CAS), start a thread with an initial prompt, then invoke `uwf thread step` repeatedly until the moderator routes to `$END`. Each step is a complete process: the moderator evaluates JSONata conditions to pick the next role, an external agent CLI produces frontmatter markdown output, and an extract pipeline validates or structures that output against the role's JSON Schema.
 
-## Monorepo Packages
+Workflow state lives entirely on disk under `~/.uncaged/workflow/`: CAS nodes for definitions and step payloads, `registry.yaml` for workflow name→hash mappings, and `threads.yaml` for active thread head pointers. Completed threads are archived to `history.jsonl`. Because there is no server process, workflows are easy to debug, fork, and inspect with ordinary CLI tools.
+
+Agents are pluggable CLI binaries (`uwf-hermes`, `uwf-builtin`, `uwf-claude-code`, or custom commands). The engine spawns the configured agent with `<thread-id>` and `<role>`, sets `UWF_EDGE_PROMPT` from the graph transition, and captures both the agent's markdown output and a detail CAS node for session replay.
+
+## Architecture
+
+Dependency layers (lower layers have no dependency on higher layers):
 
 ```
-packages/
-  workflow/                      # @uncaged/workflow — core lib (types, engine, hash, ULID, registry)
-  cli-workflow/                  # @uncaged/cli-workflow — CLI (`uncaged-workflow` command)
-  workflow-template-develop/     # @uncaged/workflow-template-develop — develop workflow template (includes roles)
-  workflow-template-solve-issue/ # @uncaged/workflow-template-solve-issue — solve-issue workflow template (includes roles)
-  workflow-agent-hermes/         # @uncaged/workflow-agent-hermes — Hermes agent adapter
-  workflow-agent-cursor/         # @uncaged/workflow-agent-cursor — Cursor agent adapter
-  workflow-agent-llm/            # @uncaged/workflow-agent-llm — LLM agent adapter
-  workflow-util-agent/           # @uncaged/workflow-util-agent — agent utilities (buildAgentPrompt, spawnCli)
+Layer 0 — Contract
+  workflow-protocol          Shared types and JSON Schema definitions
+
+Layer 1 — Shared infra
+  workflow-util              Encoding, IDs, logging, frontmatter, paths
+  workflow-moderator         JSONata graph evaluator
+
+Layer 2 — Agent framework
+  workflow-agent-kit         createAgent factory, context builder, extract pipeline
+
+Layer 3 — Agent implementations
+  workflow-agent-hermes      Hermes ACP agent (uwf-hermes)
+  workflow-agent-builtin     Built-in LLM + tools agent (uwf-builtin)
+  workflow-agent-claude-code Claude Code agent (uwf-claude-code)
+
+Layer 4 — CLI
+  cli-workflow               uwf binary — thread lifecycle, registry, CAS, setup
+
+App (uses protocol; not in the runtime engine stack)
+  workflow-dashboard         Web UI for visual workflow editing
 ```
 
-Managed with **bun workspace** using the `workspace:*` protocol.
+External CAS: [`@uncaged/json-cas`](https://www.npmjs.com/package/@uncaged/json-cas) (store API, hashing, schema validation) + `@uncaged/json-cas-fs` (filesystem backend).
+
+See [docs/architecture.md](docs/architecture.md) for the full design — three-phase engine loop, CAS node types, storage layout, agent CLI protocol, and design decisions.
+
+## Packages
+
+| Package | npm | Description | Type | README |
+|---------|-----|-------------|------|--------|
+| `cli-workflow` | `@uncaged/cli-workflow` | `uwf` CLI — thread lifecycle, workflow registry, CAS inspection, setup | cli | [README](packages/cli-workflow/README.md) |
+| `workflow-protocol` | `@uncaged/workflow-protocol` | Shared TypeScript types and JSON Schema constants | lib | [README](packages/workflow-protocol/README.md) |
+| `workflow-moderator` | `@uncaged/workflow-moderator` | JSONata graph evaluator — next role or `$END` | lib | [README](packages/workflow-moderator/README.md) |
+| `workflow-agent-kit` | `@uncaged/workflow-agent-kit` | `createAgent` factory, context builder, extract pipeline | lib | [README](packages/workflow-agent-kit/README.md) |
+| `workflow-util` | `@uncaged/workflow-util` | Crockford Base32, ULID, logger, frontmatter parsing, storage paths | lib | [README](packages/workflow-util/README.md) |
+| `workflow-agent-hermes` | `@uncaged/workflow-agent-hermes` | `uwf-hermes` — spawns Hermes chat via ACP | agent | [README](packages/workflow-agent-hermes/README.md) |
+| `workflow-agent-builtin` | `@uncaged/workflow-agent-builtin` | `uwf-builtin` — built-in LLM agent with file/shell tools | agent | [README](packages/workflow-agent-builtin/README.md) |
+| `workflow-agent-claude-code` | `@uncaged/workflow-agent-claude-code` | `uwf-claude-code` — spawns Claude Code CLI | agent | [README](packages/workflow-agent-claude-code/README.md) |
+| `workflow-dashboard` | `@uncaged/workflow-dashboard` | Web graph editor for workflow YAML (private, alpha) | app | [README](packages/workflow-dashboard/README.md) |
 
 ## Quick Start
 
 ```bash
-# Install dependencies
-bun install
+# 1. Configure provider, model, and default agent
+uwf setup
 
-# Build all packages
-bun run build
+# 2. Register a workflow from YAML
+uwf workflow put examples/solve-issue.yaml
 
-# Register a workflow bundle
-uncaged-workflow workflow add solve-issue dist/packages/workflow-template-solve-issue/solve-issue.esm.js
+# 3. Start a thread (creates head pointer; does not execute)
+uwf thread start solve-issue -p "Fix the login redirect bug"
 
-# Run a workflow
-uncaged-workflow run solve-issue --prompt "Fix bug #42"
+# 4. Execute steps (one at a time, until done)
+uwf thread step <thread-id>
 ```
 
-## CLI Usage
+Use `-c, --count <number>` on `thread step` to run multiple steps in one invocation. Override the agent with `--agent <cmd>`.
 
-```bash
-uncaged-workflow                   # Print full command usage (exits with status 1)
-uncaged-workflow workflow list     # List registered workflows
-uncaged-workflow run <name>        # Start a workflow thread
-uncaged-workflow thread list       # List all threads
-uncaged-workflow thread show <id>  # Inspect a thread
-uncaged-workflow skill             # Agent-consumable reference docs
-```
+## CLI Reference
 
-Run `uncaged-workflow` with no arguments to print usage, or `uncaged-workflow skill cli` for the full CLI skill reference.
+Global options: `-V, --version`, `--format <json|yaml>`, `-h, --help`.
+
+| Group | Commands |
+|-------|----------|
+| **thread** | `start`, `step`, `show`, `list`, `kill`, `steps`, `read`, `fork`, `step-details` |
+| **workflow** | `put`, `show`, `list` |
+| **cas** | `get`, `put`, `put-text`, `has`, `refs`, `walk`, `reindex`, `schema list`, `schema get` |
+| **setup** | Interactive or `--provider`, `--base-url`, `--api-key`, `--model`, `--agent` |
+| **skill** | `cli` — print markdown reference of all uwf commands |
+| **log** | `list`, `show`, `clean` — process-level debug logs |
+
+Config is stored in `~/.uncaged/workflow/config.yaml`. API keys go in `~/.uncaged/workflow/.env`.
+
+Detailed command usage, options, and examples: [packages/cli-workflow/README.md](packages/cli-workflow/README.md).
 
 ## Development
 
 ```bash
-bun run check    # Biome lint + format check
-bun run format   # Auto-format with Biome
-bun test         # Run tests
+bun install --no-cache     # Install dependencies
+bun run build              # tsc --build (all packages)
+bun run check              # tsc + biome + lint-log-tags
+bun run format             # Auto-format with Biome
+bun test                   # Run all tests
 ```
 
-## Architecture
-
-See [docs/architecture.md](docs/architecture.md) for the full design — three-phase engine loop, bundle contract, storage layout, and design decisions.
+Managed with **bun workspace**. See [CLAUDE.md](CLAUDE.md) for coding conventions.

biome.json

@@ -5,6 +5,8 @@
       "**",
       "!**/dist",
       "!**/node_modules",
+      "!**/legacy-packages",
+      "!scripts",
       "!packages/workflow/workflow",
       "!xiaoju/scripts/bundle.ts"
     ]
@@ -15,6 +17,15 @@
     "indentWidth": 2,
     "lineWidth": 100
   },
+  "css": {
+    "parser": {
+      "cssModules": true,
+      "tailwindDirectives": true
+    },
+    "linter": {
+      "enabled": false
+    }
+  },
   "javascript": {
     "formatter": {
       "quoteStyle": "double",
@@ -36,7 +47,7 @@
       }
     },
     {
-      "includes": ["**/*.d.ts"],
+      "includes": ["**/*.d.ts", "**/vitest.config.*"],
       "linter": {
         "rules": {
           "style": {
@@ -44,6 +55,16 @@
           }
         }
       }
+    },
+    {
+      "includes": ["**/cli.ts", "**/setup.ts"],
+      "linter": {
+        "rules": {
+          "suspicious": {
+            "noConsole": "off"
+          }
+        }
+      }
     }
   ],
   "linter": {

docs/architecture.md

@@ -85,8 +85,13 @@ description: "End-to-end issue resolution"
 roles:
   planner:
     description: "Creates implementation plan"
-    systemPrompt: "You are a planning agent. Analyze the issue and create a step-by-step plan."
-    outputSchema:
+    goal: "You are a planning agent. Analyze the issue and create a step-by-step plan."
+    capabilities:
+      - issue-analysis
+      - planning
+    procedure: "Analyze the issue and create a detailed, actionable implementation plan."
+    output: "Output the plan summary and list of concrete steps."
+    meta:
       type: object
       properties:
         plan: { type: string }
@@ -94,8 +99,13 @@ roles:
       required: [plan, steps]
   developer:
     description: "Implements code changes"
-    systemPrompt: "You are a developer agent. Implement the plan."
-    outputSchema:
+    goal: "You are a developer agent. Implement the plan."
+    capabilities:
+      - file-edit
+      - shell
+    procedure: "Implement the plan. Write code, tests, and ensure existing tests pass."
+    output: "List all files changed and provide a summary of the implementation."
+    meta:
       type: object
       properties:
         filesChanged: { type: array, items: { type: string } }
@@ -103,8 +113,12 @@ roles:
       required: [filesChanged, summary]
   reviewer:
     description: "Reviews code changes"
-    systemPrompt: "You are a code reviewer. Review the implementation."
-    outputSchema:
+    goal: "You are a code reviewer. Review the implementation."
+    capabilities:
+      - code-review
+    procedure: "Review the implementation against the plan."
+    output: "Approve or reject with detailed comments."
+    meta:
       type: object
       properties:
         approved: { type: boolean }
@@ -133,7 +147,7 @@ graph:
 
 Key properties:
 
-- **`roles`** — inline role definitions; each `outputSchema` is a JSON Schema (stored as its own CAS node on registration)
+- **`roles`** — inline role definitions; each `meta` is a JSON Schema (stored as its own CAS node on registration)
 - **`conditions`** — named JSONata expressions evaluated against the `ModeratorContext`
 - **`graph`** — `Record<Role | "$START", Transition[]>` — first matching transition wins; `condition: null` = fallback
 - **No agent binding** — agent selection is a deployment concern, configured in `config.yaml`
@@ -156,7 +170,7 @@ Each `uwf thread step` runs exactly one cycle: moderator → agent → extract.
 │   Output: raw string (frontmatter markdown)
 │
 │   Phase 3: EXTRACT
-│   Input:  raw agent output + role's outputSchema
+│   Input:  raw agent output + role's meta schema
 │   Engine: two-layer extract (frontmatter fast path → LLM fallback)
 │   Output: CasRef to structured output node
 │
@@ -213,7 +227,7 @@ Contract:
    - Parses argv
    - Loads `.env` from storage root
    - Builds `AgentContext` by walking the CAS chain from `threads.yaml` head
-   - Resolves the role's `outputSchema` and builds `outputFormatInstruction`
+   - Resolves the role's `meta` schema and builds `outputFormatInstruction`
    - Calls the agent's `run` function
    - Runs two-layer extract on the raw output
    - Writes `StepNode` to CAS (output + detail + prev link)
@@ -242,7 +256,7 @@ scope: role
 Fixed the login redirect by updating the auth middleware...
 ```
 
-The `outputFormatInstruction` (built by `buildOutputFormatInstruction` in `workflow-agent-kit`) is prepended to the role's system prompt, so the deliverable format is the first thing the agent sees. It lists the expected frontmatter fields derived from the role's JSON Schema.
+The `outputFormatInstruction` (built by `buildOutputFormatInstruction` in `workflow-agent-kit`) is prepended to the role's system prompt, so the deliverable format is the first thing the agent sees. It lists the expected frontmatter fields derived from the role's `meta` JSON Schema.
 
 ## Two-layer extract
 
@@ -253,7 +267,7 @@ Structured output extraction uses a two-layer strategy (`workflow-agent-kit`):
 1. Parse YAML frontmatter from raw agent output (`parseFrontmatterMarkdown`)
 2. Validate required fields (`validateFrontmatter`)
 3. Build a candidate object from frontmatter fields (`status`, `next`, `confidence`, `artifacts`, `scope`)
-4. `store.put()` the candidate against the role's `outputSchema`
+4. `store.put()` the candidate against the role's `meta` schema
 5. Validate with `json-cas` schema validation
 6. If valid → return `outputHash` (zero LLM cost)
 
@@ -272,7 +286,7 @@ If the fast path returns `null` (no frontmatter, invalid, or doesn't satisfy sch
 
 `workflow-agent-kit` prepends two pieces of context to the agent's system prompt:
 
-1. **Deliverable format instruction** — generated from the role's `outputSchema`, tells the agent exactly what frontmatter fields to produce and the expected format
+1. **Deliverable format instruction** — generated from the role's `meta` schema, tells the agent exactly what frontmatter fields to produce and the expected format
 2. **Scope constraint** — "Focus exclusively on YOUR role's deliverable. Do not perform actions outside your role's scope."
 
 This ensures agents produce parseable frontmatter output without requiring per-agent format knowledge.
@@ -289,8 +303,11 @@ payload:
   roles:
     planner:
       description: "Creates implementation plan"
-      systemPrompt: "You are a planning agent..."
-      outputSchema: "5GWKR8TN1V3JA"    # cas_ref → JSON Schema node
+      goal: "You are a planning agent..."
+      capabilities: [planning, issue-analysis]
+      procedure: "Analyze the issue and create a plan."
+      output: "Output the plan summary."
+      meta: "5GWKR8TN1V3JA"    # cas_ref → JSON Schema node
   conditions:
     notApproved:
       description: "Reviewer rejected"
@@ -318,7 +335,7 @@ payload:
   start: "4TNVW8KR2B3MA"      # cas_ref → StartNode
   prev: "2MXBG6PN4A8JR"       # cas_ref → previous StepNode (null for first step)
   role: "developer"
-  output: "9KRVW3TN5F1QA"     # cas_ref → structured output (validated against outputSchema)
+  output: "9KRVW3TN5F1QA"     # cas_ref → structured output (validated against meta schema)
   detail: "7BQST3VW9F2MA"     # cas_ref → execution detail (raw turns, session data)
   agent: "uwf-hermes"         # agent command used (plain string)
 ```

docs/builtin-agent-research.md

@@ -0,0 +1,779 @@
+# Built-in Role Agent 调研
+
+## 目标
+
+实现一个内置的 role agent（暂称 `uwf-builtin`），不依赖 hermes/openclaw 等外部 agent 进程。
+直接使用 workflow config 中配置的 model，自己实现 agent run loop 和关键 toolkit。
+
+---
+
+## 关键问题
+
+### Q1: Agent 接口协议
+
+现有 agent 是怎么被 CLI 调用的？输入（argv、环境变量）和输出（stdout、CAS）格式是什么？
+
+**调研要点：**
+- `cli-workflow` 里 `spawnAgent` 的完整实现
+- AgentConfig 类型定义
+- agent 进程的 exit code 约定
+- 环境变量传递（UWF_STORAGE_ROOT 等）
+
+**答案：**
+
+#### 调用链
+
+`uwf thread step` → `cmdThreadStepOnce` → moderator 求值下一 role → `resolveAgentConfig` → `spawnAgent`。
+
+#### AgentConfig 类型
+
+```146:149:packages/workflow-protocol/src/types.ts
+export type AgentConfig = {
+  command: string;
+  args: string[];
+};
+```
+
+在 `config.yaml` 的 `agents` 段注册，例如 `hermes: { command: "uwf-hermes", args: [] }`。
+
+#### spawnAgent 行为
+
+```627:653:packages/cli-workflow/src/commands/thread.ts
+function spawnAgent(agent: AgentConfig, threadId: ThreadId, role: string): CasRef {
+  const argv = [...agent.args, threadId, role];
+  let stdout: string;
+  try {
+    stdout = execFileSync(agent.command, argv, {
+      encoding: "utf8",
+      env: process.env,
+      stdio: ["ignore", "pipe", "pipe"],
+    });
+  } catch (e) {
+  // ... stderr 拼进 fail 消息
+  }
+
+  const line = stdout.trim().split("\n").pop()?.trim() ?? "";
+  if (!isCasRef(line)) {
+    fail(`agent stdout is not a valid CAS hash: ${line || "(empty)"}`);
+  }
+  return line;
+}
+```
+
+| 项目 | 约定 |
+|------|------|
+| **argv** | `[...agent.args, <thread-id>, <role>]`，即 `process.argv[2]`=threadId，`process.argv[3]`=role（与 `createAgent` 的 `parseArgv` 一致） |
+| **stdin** | 忽略 |
+| **stdout** | 纯文本，**最后一行**必须是新 `StepNode` 的 CAS hash（13 字符 Crockford Base32） |
+| **stderr** | 失败时 CLI 会附带 stderr；成功时无约定 |
+| **exit code** | `0` = 成功；非 0 时 `execFileSync` 抛错，step 失败 |
+| **环境变量** | 继承父进程 `process.env`（含 storage root、API key 等） |
+| **链头更新** | **不由 agent 负责**；agent 只写 CAS StepNode，CLI 在拿到 stdout hash 后更新 `threads.yaml` |
+
+Agent 解析优先级（`resolveAgentConfig`）：
+
+1. CLI `--agent` override（整段 command + args 字符串）
+2. `config.agentOverrides[workflow.name][role]`
+3. `config.defaultAgent`
+
+#### 环境变量：Storage Root
+
+文档中写的 `UWF_STORAGE_ROOT` **在当前代码中不存在**。实际优先级（`workflow-agent-kit` / `cli-workflow` 一致）：
+
+```33:43:packages/workflow-agent-kit/src/storage.ts
+export function resolveStorageRoot(): string {
+  const internal = process.env.UNCAGED_WORKFLOW_STORAGE_ROOT;
+  if (internal !== undefined && internal !== "") {
+    return internal;
+  }
+  const userOverride = process.env.WORKFLOW_STORAGE_ROOT;
+  if (userOverride !== undefined && userOverride !== "") {
+    return userOverride;
+  }
+  return getDefaultStorageRoot();
+}
+```
+
+Agent 子进程通过继承的 `process.env` 与父 CLI 共享同一 storage root；`createAgent` 内还会 `loadDotenv({ path: getEnvPath(storageRoot) })` 加载 `~/.uncaged/workflow/.env`。
+
+#### Agent 侧职责（设计文档 + 实现）
+
+- 读 `threads.yaml` 链头，构建 context，执行 role
+- 将 `StepNode` 写入 CAS（`output` / `detail` / `agent` / `prev` / `start`）
+- stdout 打印 step hash
+- **不**更新 `threads.yaml`
+
+---
+
+### Q2: createAgent 工厂
+
+workflow-agent-kit 的 `createAgent` 做了什么？它的完整生命周期是什么？
+
+**调研要点：**
+- `AgentOptions` 类型的 `run` 和 `continue` 回调签名
+- `AgentRunResult` 的完整定义
+- retry 逻辑（frontmatter 校验失败后的重试机制）
+- `persistStep` 写入 CAS 的 StepNode 结构
+
+**答案：**
+
+#### 类型定义
+
+```4:35:packages/workflow-agent-kit/src/types.ts
+export type AgentContext = ModeratorContext & {
+  threadId: ThreadId;
+  role: string;
+  store: Store;
+  workflow: WorkflowPayload;
+  outputFormatInstruction: string;
+};
+
+export type AgentRunResult = {
+  output: string;
+  detailHash: CasRef;
+  sessionId: string;
+};
+
+export type AgentContinueFn = (
+  sessionId: string,
+  message: string,
+  store: AgentContext["store"],
+) => Promise<AgentRunResult>;
+
+export type AgentRunFn = (ctx: AgentContext) => Promise<AgentRunResult>;
+
+export type AgentOptions = {
+  name: string;
+  run: AgentRunFn;
+  continue: AgentContinueFn;
+};
+```
+
+- **`run(ctx)`**：首次执行，返回原始 agent 文本 `output`、审计用 `detailHash`、用于续聊的 `sessionId`。
+- **`continue(sessionId, message, store)`**：在同一 session 上追加用户消息（用于 frontmatter 纠错），再次返回 `AgentRunResult`。
+
+`createAgent(options)` 返回 `() => Promise<void>`，作为 agent CLI 的 `main`（见 `uwf-hermes` 的 `cli.ts`）。
+
+#### 生命周期（按执行顺序）
+
+```101:152:packages/workflow-agent-kit/src/run.ts
+export function createAgent(options: AgentOptions): () => Promise<void> {
+  return async function main(): Promise<void> {
+    const { threadId, role } = parseArgv(process.argv);
+    const storageRoot = resolveStorageRoot();
+    loadDotenv({ path: getEnvPath(storageRoot) });
+
+    const ctx = await buildContextWithMeta(threadId, role);
+    // 1. 校验 role 存在
+    // 2. 从 CAS 取 frontmatter JSON Schema → buildOutputFormatInstruction → ctx.outputFormatInstruction
+
+    let agentResult = await options.run(ctx);
+
+    let outputHash = await tryExtractOutput(agentResult.output, roleDef.frontmatter, ctx);
+
+    for (let retry = 0; retry < MAX_FRONTMATTER_RETRIES && outputHash === null; retry++) {
+      const correctionMessage = "Your previous response did not contain valid YAML frontmatter...";
+      agentResult = await options.continue(agentResult.sessionId, correctionMessage, ctx.meta.store);
+      outputHash = await tryExtractOutput(agentResult.output, roleDef.frontmatter, ctx);
+    }
+
+    if (outputHash === null) { fail(...); }
+
+    const stepHash = await persistStep({ ctx, outputHash, detailHash: agentResult.detailHash, agentName });
+    process.stdout.write(`${stepHash}\n`);
+  };
+}
+```
+
+| 阶段 | 行为 |
+|------|------|
+| 解析 argv | `argv[2]=threadId`, `argv[3]=role`，缺失则 `stderr` + `exit(1)` |
+| Context | `buildContextWithMeta` + 可选 `outputFormatInstruction` |
+| Run | `options.run(ctx)` |
+| Extract | **仅** `tryFrontmatterFastPath`（见 Q4）；**不**调用 `extract()` LLM fallback |
+| Retry | 最多 `MAX_FRONTMATTER_RETRIES = 2` 次 `continue` + 再试 fast-path |
+| Persist | `persistStep` → `writeStepNode` |
+| 输出 | stdout 一行 step CAS hash |
+
+#### StepNode 写入结构
+
+```44:68:packages/workflow-agent-kit/src/run.ts
+async function writeStepNode(options: {
+  store: AgentStore["store"];
+  schemas: AgentStore["schemas"];
+  startHash: CasRef;
+  prevHash: CasRef | null;
+  role: string;
+  outputHash: CasRef;
+  detailHash: CasRef;
+  agentName: string;
+}): Promise<CasRef> {
+  const payload: StepNodePayload = {
+    start: options.startHash,
+    prev: options.prevHash,
+    role: options.role,
+    output: options.outputHash,
+    detail: options.detailHash,
+    agent: options.agentName,
+  };
+  // store.put(stepNode schema) + validate
+}
+```
+
+`agentName` 经 `agentLabel(name)` 规范化：已有 `uwf-` 前缀则原样，否则加 `uwf-`（如 `hermes` → `uwf-hermes`）。
+
+`prevHash`：若链头仍是 `StartNode` 则为 `null`，否则为当前 head step hash。
+
+---
+
+### Q3: Context Builder
+
+`buildContextWithMeta` 构建了什么上下文给 agent？
+
+**调研要点：**
+- `AgentContext` 完整类型定义（所有字段）
+- context 构建过程（CAS chain walk）
+- `outputFormatInstruction` 怎么生成的
+- role definition 怎么获取（从 workflow YAML）
+
+**答案：**
+
+#### AgentContext 字段
+
+继承 `ModeratorContext`：
+
+```60:68:packages/workflow-protocol/src/types.ts
+export type ModeratorContext = {
+  start: StartNodePayload;
+  steps: StepContext[];
+};
+```
+
+```48:51:packages/workflow-protocol/src/types.ts
+export type StartNodePayload = {
+  workflow: CasRef;
+  prompt: string;
+};
+```
+
+```61:63:packages/workflow-protocol/src/types.ts
+export type StepContext = Omit<StepRecord, "output"> & {
+  output: unknown;
+};
+```
+
+`AgentContext` 额外字段：
+
+| 字段 | 类型 | 含义 |
+|------|------|------|
+| `threadId` | `ThreadId` | 当前线程 |
+| `role` | `string` | 本步要执行的角色名 |
+| `store` | `Store` | CAS store（读写节点） |
+| `workflow` | `WorkflowPayload` | 已从 CAS 加载的 workflow 定义 |
+| `outputFormatInstruction` | `string` | 由 `createAgent` 根据 role 的 frontmatter schema 生成；`buildContext*` 初始为 `""` |
+
+`buildContextWithMeta` 还返回 `meta`：
+
+```148:154:packages/workflow-agent-kit/src/context.ts
+export type BuildContextMeta = {
+  storageRoot: string;
+  store: Store;
+  schemas: AgentStore["schemas"];
+  headHash: CasRef;
+  chain: ChainState;
+};
+```
+
+#### CAS chain walk
+
+1. 从 `threads.yaml[threadId]` 取 `headHash`
+2. `walkChain`：若 head 是 `StartNode`，`stepsNewestFirst=[]`；否则沿 `prev` 收集所有 `StepNode`， newest-first
+3. `buildHistory`：反转为时间序，`expandOutput` 把每步 `output` CasRef 展开为 JSON payload（供 prompt / JSONata 使用）
+4. `loadWorkflow`：从 `start.workflow` CasRef 加载 `WorkflowPayload`
+
+#### Role definition 来源
+
+- 作者写在 workflow YAML 的 `roles.<name>`（`goal`, `capabilities`, `procedure`, `output`, `frontmatter` 等）
+- `uwf workflow put` 时 `frontmatter` 内联 JSON Schema 经 `putSchema` 存入 CAS，workflow 里存的是 **CasRef**
+- Agent 运行时：`ctx.workflow.roles[ctx.role]` → `RoleDefinition`
+
+#### outputFormatInstruction
+
+在 `createAgent` 中，若 `getSchema(store, roleDef.frontmatter)` 非空，则：
+
+```typescript
+ctx.outputFormatInstruction = buildOutputFormatInstruction(frontmatterSchema);
+```
+
+`buildOutputFormatInstruction` 根据 JSON Schema 的 `properties` 生成「必须以 `---` YAML frontmatter 开头」的说明和示例字段列表（见 `build-output-format-instruction.ts`）。
+
+各 agent 实现（Hermes / Claude Code）在组装 prompt 时把该块放在最前，再接 `buildRolePrompt(roleDef)`。
+
+---
+
+### Q4: Extract Pipeline
+
+agent 输出怎么被处理成结构化数据？
+
+**调研要点：**
+- frontmatter fast-path 的完整逻辑
+- LLM extract fallback 的实现（`extract.ts`）
+- frontmatter schema 从哪里来（role 定义里的 `frontmatter` 字段）
+- 校验失败时的 correction prompt 是什么
+
+**答案：**
+
+#### Schema 来源
+
+Workflow YAML 中每个 role 的 `frontmatter:` 段是 JSON Schema 对象；注册时：
+
+```66:76:packages/cli-workflow/src/commands/workflow.ts
+async function resolveFrontmatterRef(..., frontmatter: unknown): Promise<CasRef> {
+  // 校验为 JSON Schema → putSchema → 返回 CasRef
+}
+```
+
+运行时 `roleDef.frontmatter` 即该 schema 的 CAS hash；structured `output` 节点用**同一 schema** 写入 CAS。
+
+#### Frontmatter fast-path（createAgent 实际使用的路径）
+
+```148:195:packages/workflow-agent-kit/src/frontmatter.ts
+export async function tryFrontmatterFastPath(
+  raw: string,
+  outputSchema: CasRef,
+  store: Store,
+): Promise<FrontmatterFastPathResult | null>
+```
+
+流程：
+
+1. `parseFrontmatterMarkdown(raw)` → 标准 agent 字段（`status`, `next`, `confidence`, `artifacts`, `scope`）+ body
+2. `validateFrontmatter` 失败 → `null`
+3. `getSchema(store, outputSchema)` + `extractSchemaFields` 得到 role 需要的属性名
+4. `buildCandidate`：从标准 frontmatter + YAML 原始字段拼出符合 schema 的对象
+5. `store.put(outputSchema, candidate)` + `validate` → 成功则 `{ body, outputHash }`
+
+**永不抛错**，失败返回 `null`。
+
+#### LLM extract fallback（已实现但未接入 createAgent）
+
+```135:181:packages/workflow-agent-kit/src/extract.ts
+export async function extract(
+  rawOutput: string,
+  outputSchema: CasRef,
+  config: WorkflowConfig,
+): Promise<ExtractResult>
+```
+
+- 模型：`resolveExtractModelAlias(config)` → `modelOverrides.extract` → `models.extract` → `models.default` → `defaultModel`
+- HTTP：`POST {baseUrl}/chat/completions`，`response_format: { type: "json_object" }`
+- System：要求按 JSON Schema 从 agent 输出提取单个 JSON 对象
+- 校验通过后 `store.put(outputSchema, structured)`
+
+**重要：`createAgent` 当前未调用 `extract()`**。fast-path 失败且 2 次 `continue` 仍失败则直接 `fail()`。builtin agent 若希望无 frontmatter 也能跑，需在 kit 或 builtin 层显式接入 `extract()`。
+
+#### Correction prompt（retry）
+
+```125:128:packages/workflow-agent-kit/src/run.ts
+const correctionMessage =
+  "Your previous response did not contain valid YAML frontmatter matching the role schema.\n" +
+  "You MUST begin your response with a YAML frontmatter block (--- delimited).\n" +
+  "Please output ONLY the corrected frontmatter block followed by your work.";
+```
+
+通过 `options.continue(sessionId, correctionMessage, store)` 发给外部 agent；builtin 需在自有 message 历史里 append 同等语义的 user 消息。
+
+---
+
+### Q5: Model 配置与 LLM 调用
+
+workflow 怎么配置和使用 model？
+
+**调研要点：**
+- `WorkflowConfig` 中 providers/models/defaultModel/modelOverrides 的完整定义
+- `resolveModel` 函数的实现
+- `chatCompletionText` 的实现（OpenAI 兼容 HTTP 客户端）
+- 有没有 streaming 支持？tool calling 支持？
+
+**答案：**
+
+#### WorkflowConfig
+
+```136:160:packages/workflow-protocol/src/types.ts
+export type ProviderConfig = {
+  baseUrl: string;
+  apiKeyEnv: string;
+};
+
+export type ModelConfig = {
+  provider: ProviderAlias;
+  name: string;
+};
+
+export type WorkflowConfig = {
+  providers: Record<ProviderAlias, ProviderConfig>;
+  models: Record<ModelAlias, ModelConfig>;
+  agents: Record<AgentAlias, AgentConfig>;
+  defaultAgent: AgentAlias;
+  agentOverrides: Record<WorkflowName, Record<RoleName, AgentAlias>> | null;
+  defaultModel: ModelAlias;
+  modelOverrides: Record<Scenario, ModelAlias> | null;
+};
+```
+
+示例见 `docs/architecture.md`（`providers` / `models` / `defaultModel` / `modelOverrides.extract`）。
+
+#### resolveModel
+
+```32:50:packages/workflow-agent-kit/src/extract.ts
+export function resolveModel(config: WorkflowConfig, alias: ModelAlias): ResolvedLlmProvider {
+  const modelEntry = config.models[alias];
+  const providerEntry = config.providers[modelEntry.provider];
+  const apiKey = process.env[providerEntry.apiKeyEnv];
+  return { baseUrl: providerEntry.baseUrl, apiKey, model: modelEntry.name };
+}
+```
+
+`ResolvedLlmProvider = { baseUrl, apiKey, model }`。
+
+Extract 专用别名解析：
+
+```18:30:packages/workflow-agent-kit/src/extract.ts
+export function resolveExtractModelAlias(config: WorkflowConfig): ModelAlias {
+  return config.modelOverrides?.extract ?? (config.models.extract ? "extract" : config.models.default ? "default" : config.defaultModel);
+}
+```
+
+**尚无** `modelOverrides` 按 role/workflow 解析 agent 主模型的函数；builtin 首版可用 `config.defaultModel`，扩展时可加 `modelOverrides.agent` 或与 `agentOverrides` 对称的表。
+
+#### chatCompletionText
+
+```87:124:packages/workflow-agent-kit/src/extract.ts
+async function chatCompletionText(
+  provider: ResolvedLlmProvider,
+  messages: Array<{ role: "system" | "user"; content: string }>,
+): Promise<string>
+```
+
+| 能力 | 现状 |
+|------|------|
+| 协议 | OpenAI 兼容 `POST /chat/completions` |
+| Streaming | **无**（一次性 `response.text()`） |
+| Tool calling | **无**（无 `tools` / `tool_calls` 字段） |
+| 多模态 | **无**（仅 text `content`） |
+| Extract 专用 | `response_format: { type: "json_object" }` |
+
+builtin agent 的 run loop 需要**新写**带 `tools` 的 completion 客户端（可放在 `workflow-agent-builtin` 或扩展 `workflow-agent-kit` 的 `llm/` 模块），不能复用当前 `chatCompletionText` 而不改。
+
+---
+
+### Q6: Hermes Agent 参考实现
+
+`uwf-hermes` 是怎么实现 `run` 和 `continue` 的？
+
+**调研要点：**
+- prompt 怎么组装的（outputFormatInstruction + rolePrompt + task + history）
+- hermes CLI 的调用参数
+- session management（resume）
+- 输出怎么捕获
+
+**答案：**
+
+#### Prompt 组装
+
+```40:53:packages/workflow-agent-hermes/src/hermes.ts
+export function buildHermesPrompt(ctx: AgentContext): string {
+  const roleDef = ctx.workflow.roles[ctx.role];
+  const rolePrompt = roleDef !== undefined ? buildRolePrompt(roleDef) : "";
+  const parts: string[] = [];
+  if (ctx.outputFormatInstruction !== "") {
+    parts.push(ctx.outputFormatInstruction, "");
+  }
+  parts.push(rolePrompt, "", "## Task", ctx.start.prompt);
+  const historyBlock = buildHistorySummary(ctx.steps);
+  if (historyBlock !== "") {
+    parts.push("", historyBlock);
+  }
+  return parts.join("\n");
+}
+```
+
+`buildRolePrompt` 生成 `## Goal` / `## Capabilities` / `## Prepare`（含 `generateCliReference()`）/ `## Procedure` / `## Output`。
+
+`buildHistorySummary`：每步 `role`、`JSON.stringify(step.output)`、`agent`。
+
+Hermes 把**整段 prompt 作为单条 user 消息**传给 `hermes chat -q`（无独立 system channel）。
+
+#### Hermes CLI 参数
+
+首次：
+
+```88:97:packages/workflow-agent-hermes/src/hermes.ts
+spawnHermes(["chat", "-q", prompt, "--yolo", "--max-turns", "90", "--quiet"]);
+```
+
+续聊：
+
+```100:114:packages/workflow-agent-hermes/src/hermes.ts
+spawnHermes(["chat", "--resume", sessionId, "-q", message, "--yolo", "--max-turns", "90", "--quiet"]);
+```
+
+#### Session
+
+- stdout/stderr 中解析 `session_id: <id>`（`parseSessionIdFromStdout`）
+- 会话文件：`~/.hermes/sessions/session_<id>.json`
+- `loadHermesSession` → `storeHermesSessionDetail`：每 assistant/tool 消息写成 CAS turn 节点，汇总为 `detail`；**output 文本** = 最后一条非空 `assistant` 的 `content`
+
+#### 与 createAgent 的衔接
+
+```157:164:packages/workflow-agent-hermes/src/hermes.ts
+export function createHermesAgent(): () => Promise<void> {
+  return createAgent({ name: "hermes", run: runHermes, continue: continueHermes });
+}
+```
+
+`uwf-hermes` 入口：`createHermesAgent()` 即 main。
+
+Claude Code 包（`workflow-agent-claude-code`）结构相同：`buildClaudeCodePrompt` 同构，`claude -p` + `--resume` + JSON stdout 解析。
+
+---
+
+### Q7: Toolkit 需求分析
+
+要实现一个自给自足的 agent，最少需要哪些 tool？
+
+**调研要点：**
+- 现有 workflow example（solve-issue.yaml）里 role 都做什么任务
+- hermes agent 在 workflow 场景下常用哪些 tool
+- 哪些 tool 是 agent loop 必须的（如 file read/write、shell exec、web fetch）
+
+**答案：**
+
+#### solve-issue.yaml 角色能力
+
+| Role | capabilities | 隐含需求 |
+|------|----------------|----------|
+| planner | issue-analysis, planning | 读上下文/仓库、总结，通常不需写代码 |
+| developer | file-edit, shell, testing | **读文件、写文件、执行命令** |
+| reviewer | code-review, static-analysis | 读 diff/文件、静态分析（可读+可选 shell） |
+
+#### Hermes 侧
+
+Hermes 自带完整 agent runtime（`--yolo`、max-turns），tool 集由 Hermes 项目定义，workflow 不配置。从 session JSON 可见 `tool_calls` 被记入 detail，常见包括文件与 shell 类工具。
+
+#### Builtin 最小 toolkit 建议
+
+| 优先级 | Tool | 用途 |
+|--------|------|------|
+| P0 | `read_file` | 读仓库/配置/issue 上下文 |
+| P0 | `write_file` / `edit_file` | developer 改代码 |
+| P0 | `run_command` | 测试、构建、git（需 cwd + timeout + 输出截断） |
+| P1 | `list_dir` / `glob` | 导航代码库 |
+| P1 | `grep` | 搜索符号/引用 |
+| P2 | `fetch_url` | 查文档（planner 偶尔需要） |
+
+**不需要**在 builtin 里实现 moderator / workflow 路由工具——仍由 `uwf thread step` + JSONata 负责。
+
+#### Agent loop 必须能力
+
+1. 多轮 LLM 调用 + **OpenAI-style tool_calls** 解析与执行
+2. 将 tool 结果 append 回 messages
+3. 终止条件：模型不再请求 tool，或达到 `maxTurns`
+4. 最终响应须含合法 YAML frontmatter（满足 Q4），供 `createAgent` fast-path
+
+---
+
+## 方案草案
+
+（调研完成后基于以上答案撰写）
+
+### 架构设计
+
+```mermaid
+flowchart TB
+  subgraph cli ["cli-workflow"]
+    Step["uwf thread step"]
+    Spawn["spawnAgent(uwf-builtin, threadId, role)"]
+    Step --> Spawn
+  end
+
+  subgraph builtin_pkg ["@uncaged/workflow-agent-builtin"]
+    Main["createBuiltinAgent() = createAgent({...})"]
+    Prompt["buildBuiltinPrompt(ctx)"]
+    Loop["runBuiltinLoop(provider, messages, tools)"]
+    Tools["Toolkit: read/write/exec/..."]
+    Detail["storeBuiltinDetail(turns)"]
+    Main --> Prompt
+    Main --> Loop
+    Loop --> Tools
+    Loop --> Detail
+  end
+
+  subgraph kit ["workflow-agent-kit"]
+    Ctx["buildContextWithMeta"]
+    FM["tryFrontmatterFastPath"]
+    Persist["persistStep"]
+    Ctx --> Main
+    Main --> FM
+    FM --> Persist
+  end
+
+  subgraph cas ["CAS / config"]
+    Config["config.yaml models/providers"]
+    CAS["cas/ + threads.yaml"]
+  end
+
+  Spawn --> Main
+  Config --> Loop
+  CAS --> Ctx
+  Persist --> CAS
+  Spawn -->|"stdout: step hash"| Step
+```
+
+**新包**：`packages/workflow-agent-builtin`，bin `uwf-builtin`，仅依赖 `workflow-agent-kit`、`workflow-protocol`、`workflow-util`（可选 `@uncaged/json-cas` 写 detail schema）。
+
+**分层**：
+
+| 层 | 职责 |
+|----|------|
+| `createAgent`（kit） | argv、context、frontmatter extract、StepNode、stdout 协议 — **不变** |
+| `builtin/agent.ts` | `run` / `continue` 实现 |
+| `builtin/llm.ts` | OpenAI 兼容 chat + tools（可后续抽到 kit） |
+| `builtin/tools/*.ts` | 各 tool 的 JSON Schema + handler |
+| `builtin/prompt.ts` | 复用 Hermes 的 prompt 拼接逻辑（或抽到 kit 的 `buildAgentPrompt`） |
+| `builtin/detail.ts` | 类似 Hermes：每轮 assistant/tool 写入 CAS detail |
+
+**配置集成**：
+
+```yaml
+agents:
+  builtin:
+    command: "uwf-builtin"
+    args: []
+defaultAgent: "builtin"   # 或 agentOverrides 按 role 指定
+```
+
+模型：首版 `resolveModel(config, config.defaultModel)`；后续可增加 `modelOverrides.agent` 或 per-role 映射。
+
+---
+
+### Agent Run Loop
+
+伪代码（单次 `run(ctx)`）：
+
+```
+1. provider ← resolveModel(loadWorkflowConfig(), defaultModel)
+2. system ← buildBuiltinPrompt(ctx)   // outputFormatInstruction + buildRolePrompt + Task + History
+3. messages ← [{ role: "system", content: system }]
+4. sessionId ← newULID()              // 内存或临时目录，供 continue 使用
+5. turns ← []
+
+6. for turn in 1..MAX_TURNS:
+     response ← chatCompletionWithTools(provider, messages, TOOL_DEFINITIONS)
+     record assistant message + tool_calls in turns
+
+     if response has no tool_calls:
+       finalText ← response.content
+       break
+
+     for each tool_call:
+       result ← executeTool(tool_call, { cwd: process.cwd() })
+       messages.push tool result
+       record in turns
+
+7. if no finalText with valid frontmatter after loop:
+     optionally one-shot "finalize" message without tools
+
+8. detailHash ← storeBuiltinDetail(store, sessionId, turns, metadata)
+9. return { output: finalText, detailHash, sessionId }
+```
+
+**`continue(sessionId, message, store)`**：
+
+- 从内存/磁盘恢复 `messages` + `turns`
+- `messages.push({ role: "user", content: message })`（correction 或续聊）
+- 从步骤 6 继续，步数上限可单独设小一点（如 3）
+- 返回新的 `AgentRunResult`
+
+**与 frontmatter 的配合**：
+
+- system prompt 已含 `outputFormatInstruction`；最后一轮可强制 user：`Now output your final answer with YAML frontmatter only if you have not yet.`
+- 仍依赖 `createAgent` 的 fast-path + 最多 2 次 continue
+
+**安全**：
+
+- `run_command`：白名单或需 `UWF_BUILTIN_ALLOW_SHELL=1`，默认工作区限定在 `process.cwd()` 或 `start` 中将来扩展的 `workspace` 字段
+- 路径：禁止 `..` 逃逸出 workspace root
+
+---
+
+### Toolkit 设计
+
+统一注册表：
+
+```typescript
+type BuiltinTool = {
+  name: string;
+  description: string;
+  parameters: JSONSchema; // object type
+  execute: (args: unknown, ctx: ToolContext) => Promise<string>;
+};
+
+type ToolContext = {
+  cwd: string;
+  storageRoot: string;
+};
+```
+
+| Tool name | OpenAI function | 行为摘要 |
+|-----------|-----------------|----------|
+| `read_file` | `read_file` | `{ path }` → UTF-8 文本，大小上限 |
+| `write_file` | `write_file` | `{ path, content }` → 写盘，返回确认 |
+| `edit_file` | 可选 | search/replace 块，减少 token |
+| `run_command` | `run_command` | `{ command, cwd? }` → stdout/stderr 截断 |
+| `list_dir` | `list_dir` | `{ path }` → 条目列表 |
+| `grep` | `grep` | `{ pattern, path? }` → 匹配行 |
+
+**LLM 请求形状**（扩展 extract 客户端）：
+
+```json
+{
+  "model": "...",
+  "messages": [...],
+  "tools": [{ "type": "function", "function": { "name", "description", "parameters" } }],
+  "tool_choice": "auto"
+}
+```
+
+解析 `choices[0].message.tool_calls`，执行后以 `{ role: "tool", tool_call_id, content }` 回传。
+
+**不提供** streaming 首版；detail CAS 记录每轮 tool 名/参数/结果摘要供 `uwf thread step-details` 调试。
+
+---
+
+### 与现有架构的集成
+
+| 集成点 | 方式 |
+|--------|------|
+| CLI 协议 | 实现标准 agent CLI：`uwf-builtin <thread-id> <role>`，stdout 一行 step hash，exit 0/1 |
+| 工厂 | `export function createBuiltinAgent()` → `createAgent({ name: "builtin", run, continue })` |
+| Context / Prompt | 复用 `buildContextWithMeta`、`buildRolePrompt`、`buildOutputFormatInstruction`；prompt 布局对齐 `buildHermesPrompt` |
+| 结构化输出 | 优先 YAML frontmatter fast-path；可选后续在 `createAgent` 增加 `extract()` fallback 开关 |
+| 配置 | `config.yaml` 增加 `agents.builtin`；`uwf setup` 可选默认 agent |
+| 存储 | `resolveStorageRoot()` + `loadWorkflowConfig` + `getEnvPath`；与 Hermes 相同，**不**改 `threads.yaml` 写入方 |
+| 测试 | 单元测试：tool handlers、prompt 组装、mock LLM tool loop；集成测试：临时 storage root + fake provider |
+| 发布 | 新包 `@uncaged/workflow-agent-builtin`，bin `uwf-builtin`，加入 `scripts/publish-all.mjs` |
+
+**明确不做**：
+
+- 不替代 moderator / 不在 agent 内调用 `uwf thread step`
+- 不依赖 Hermes/OpenClaw/Claude Code 二进制
+- 首版不实现 streaming、不实现 MCP
+
+**建议实现顺序**：
+
+1. `llm.ts`：tool calling HTTP 客户端 + 单测
+2. P0 tools + `runBuiltinLoop`
+3. `createBuiltinAgent` + detail CAS
+4. `config` / docs / `examples` 可选 `agentOverrides` 演示
+5. （可选）`createAgent` 接入 `extract()` fallback

docs/investigations/issue-418-acp-resume.md

@@ -0,0 +1,73 @@
+# Issue #418: ACP session/resume 返回空文本
+
+## 调研日期: 2026-05-23
+
+## 根因
+
+`session/resume` 在 restore 路径下 `_make_agent()` 失败，异常被静默吞掉。
+
+### 完整调用链
+
+```
+resume_session(sid)
+  → update_cwd(sid)
+    → get_session(sid) → _restore(sid)
+      → _make_agent()
+        → resolve_runtime_provider("custom") 失败（line 548-561）
+        → AIAgent() 抛出 "No LLM provider configured"（line 564）
+      → except Exception 静默吞掉（line 482-484）→ return None
+    → return None
+  → state is None → fallback: create_session()（新 sid，无历史）
+```
+
+### 关键代码位置（acp_adapter/session.py）
+
+- `_restore()` line 426-498: 从 DB 恢复 session，但 except 太宽泛
+- `_make_agent()` line 520-568: provider 解析在 restore 路径下不完整
+- Line 548-561: `resolve_runtime_provider("custom")` 失败后，`base_url` 虽然从 DB 取到了但没传给 AIAgent
+
+### 实测行为
+
+1. Phase 1: `session/new` + `prompt` → 正常，有 `agent_message_chunk`
+2. Phase 2: `session/resume` + `prompt`
+   - resume 返回成功，但 `available_commands_update` 里 sessionId 是新的（create_session fallback）
+   - 用原始 sid 发 prompt → `stopReason: "refusal"`（session 不在内存中）
+   - 用新 sid 发 prompt → 能跑但无历史（agent 回答"不知道 secret code"）
+
+### 验证脚本
+
+```python
+# 直接调用 _restore 验证
+cd ~/.hermes/hermes-agent
+python3 -c "
+import sys; sys.path.insert(0, '.')
+from acp_adapter.session import SessionManager
+sm = SessionManager()
+result = sm._restore('SESSION_ID_HERE')
+print(result)  # None — _make_agent 抛异常被吞掉
+"
+```
+
+### 两个 bug
+
+1. **`_make_agent` provider fallback 不完整**: restore 时 DB 里有 `base_url` 和 `api_mode`，但 `resolve_runtime_provider` 失败后这些值没被正确传递给 AIAgent
+2. **`_restore` 的 except 太宽泛**: 静默吞掉所有异常，连 warning 都只在 debug 级别，导致 resume 失败完全无感知
+
+### Hermes 版本
+
+- v0.10.0 (2026.4.16) — 初始测试
+- v0.14.0 (2026.5.16) — 更新后重新测试，bug 仍在
+- 代码路径: ~/.hermes/hermes-agent/acp_adapter/session.py
+
+### v0.14.0 测试结果 (2026-05-23)
+
+- `_restore` 仍因 `custom` provider 解析失败返回 None
+- 日志更清晰了：`WARNING: Failed to recreate agent for ACP session ...`
+- resume fallback 创建新 session（新 sid），但 agent 居然能回答之前的问题（可能通过 memory/session search）
+- 核心问题不变：sessionId 变了，client 用旧 sid 发 prompt → refusal
+
+### 上游 Issue
+
+- https://github.com/NousResearch/hermes-agent/issues/13489 — 已评论根因分析
+- https://github.com/NousResearch/hermes-agent/issues/8083 — resume 静默创建新 session
+- https://github.com/NousResearch/hermes-agent/issues/18452 — _make_agent fallback 不完整

docs/wf-stateless-design.md

@@ -112,8 +112,8 @@ uwf-hermes <thread-id> <role>
 
 **约定：**
 - `uwf step` 负责 moderator 决策，将 role 传给 agent CLI
-- agent-kit 根据 thread + role 从 CAS 读 systemPrompt / outputSchema
-- agent-kit 组装完整 prompt（role systemPrompt + thread context + user prompt from StartNode）
+- agent-kit 根据 thread + role 从 CAS 读 goal / capabilities / procedure / output / meta
+- agent-kit 组装完整 prompt（role goal/capabilities/procedure/output + thread context + user prompt from StartNode）
 - agent 执行实际逻辑，agent-kit 负责 extract
 - agent 将 StepNode 写入 CAS（含 output、detail、agent、prev），但**不挪链头指针**
 - stdout 输出新 StepNode 的 CAS hash（纯文本，一行）
@@ -143,7 +143,7 @@ uwf-hermes <thread-id> <role>
 
 #### `Workflow`
 
-Roles 和 moderator 内联在 Workflow 中，只有 outputSchema 独立为 CAS 节点（方便 json-cas 校验）。
+Roles 和 moderator 内联在 Workflow 中，只有 meta 独立为 CAS 节点（方便 json-cas 校验）。
 
 ```yaml
 type: <workflow-schema-hash>
@@ -153,16 +153,25 @@ payload:
   roles:
     planner:
       description: "Creates implementation plan"
-      systemPrompt: "You are a planning agent..."
-      outputSchema: "5GWKR8TN1V3JA"    # cas_ref → JSON Schema 节点（json-cas 内置）
+      goal: "You are a planning agent..."
+      capabilities: [planning, issue-analysis]
+      procedure: "Analyze the issue and create a plan."
+      output: "Output the plan summary."
+      meta: "5GWKR8TN1V3JA"    # cas_ref → JSON Schema 节点（json-cas 内置）
     developer:
       description: "Implements code changes"
-      systemPrompt: "You are a developer agent..."
-      outputSchema: "8CNWT4KR6D1HV"    # cas_ref → JSON Schema 节点
+      goal: "You are a developer agent..."
+      capabilities: [file-edit, shell]
+      procedure: "Implement the plan."
+      output: "List all files changed."
+      meta: "8CNWT4KR6D1HV"    # cas_ref → JSON Schema 节点
     reviewer:
       description: "Reviews code changes"
-      systemPrompt: "You are a code reviewer..."
-      outputSchema: "1VPBG9SM5E7WK"    # cas_ref → JSON Schema 节点
+      goal: "You are a code reviewer..."
+      capabilities: [code-review]
+      procedure: "Review the implementation."
+      output: "Approve or reject with comments."
+      meta: "1VPBG9SM5E7WK"    # cas_ref → JSON Schema 节点
   conditions:
     needsClarification:
       description: "Planner requests clarification from user"
@@ -189,7 +198,7 @@ payload:
         condition: null
 ```
 
-- `roles` — 内联定义，每个 role 的 `outputSchema` 是独立的 cas_ref（指向 json-cas 内置 JSON Schema 节点）
+- `roles` — 内联定义，每个 role 的 `meta` 是独立的 cas_ref（指向 json-cas 内置 JSON Schema 节点）
 - `conditions` — `Record<Name, JSONata>`，命名条件，方便画图描述
 - `graph` — `Record<Role | "$START", Transition[]>`，每个 Transition = `{ role, condition }`
 - `condition` 引用 conditions 中的 key，`null` = fallback
@@ -234,14 +243,14 @@ payload:
   start: "4TNVW8KR2B3MA"          # cas_ref → StartNode（每个 step 都引用）
   prev: "2MXBG6PN4A8JR"           # cas_ref → 前一个 StepNode，第一步为 null
   role: "developer"
-  output: "9KRVW3TN5F1QA"         # cas_ref → 结构化输出节点（符合 role 的 outputSchema）
+  output: "9KRVW3TN5F1QA"         # cas_ref → 结构化输出节点（符合 role 的 meta schema）
   detail: "7BQST3VW9F2MA"         # cas_ref → 执行详情（content node / 子 workflow terminal StepNode / ...）
   agent: "uwf-cursor"              # 实际使用的 agent 命令（纯字符串）
 ```
 
 - `start` — 每个 StepNode 都直接引用 StartNode，方便随机访问
 - `prev` — 前一个 StepNode 的 cas_ref，第一步为 `null`（不指向 StartNode）
-- `output` — cas_ref，指向符合 role outputSchema 的 CAS 节点，可用 json-cas 校验
+- `output` — cas_ref，指向符合 role meta schema 的 CAS 节点，可用 json-cas 校验
 - `detail` — cas_ref，指向执行详情。可以是原始 agent 输出（content node），也可以是子 workflow thread 的 terminal StepNode（workflowAsAgent 场景）
 - `agent` — 纯字符串，不是 CAS 节点
 
@@ -372,7 +381,7 @@ type ThreadId = string;
 /** 一个 step 的核心数据，被 StepNode payload 和 JSONata 上下文共享 */
 type StepRecord = {
   role: string;
-  output: CasRef;                    // cas_ref → 结构化输出节点（符合 role outputSchema）
+  output: CasRef;                    // cas_ref → 结构化输出节点（符合 role meta schema）
   detail: CasRef;                    // cas_ref → 执行详情（content node / 子 workflow terminal StepNode）
   agent: string;                     // 实际使用的 agent 命令（纯字符串）
 };
@@ -383,8 +392,11 @@ type StepRecord = {
 ```typescript
 type RoleDefinition = {
   description: string;
-  systemPrompt: string;
-  outputSchema: CasRef;              // cas_ref → json-cas 内置 JSON Schema 节点
+  goal: string;
+  capabilities: string[];
+  procedure: string;
+  output: string;
+  meta: CasRef;                      // cas_ref → json-cas 内置 JSON Schema 节点
 };
 
 type Transition = {

examples/analyze-topic.yaml

@@ -0,0 +1,43 @@
+name: "analyze-topic"
+description: "Single-role topic analysis using four-phase role description"
+roles:
+  analyst:
+    description: "Analyzes a given topic and produces a structured summary"
+    goal: |
+      You are a research analyst with expertise in breaking down complex topics
+      into clear, structured summaries. You think critically and cite key points.
+    capabilities:
+      - research
+      - critical-thinking
+      - structured-writing
+    procedure: |
+      Analyze the topic by:
+      1. Identifying the main thesis or question
+      2. Listing 3-5 key points with brief explanations
+      3. Noting any counterarguments or caveats
+      Keep your analysis concise (under 500 words).
+    output: |
+      Provide your analysis as markdown under the frontmatter.
+      The frontmatter must include your structured findings.
+    frontmatter:
+      type: object
+      properties:
+        thesis:
+          type: string
+        keyPoints:
+          type: array
+          items:
+            type: string
+        caveats:
+          type: string
+      required: [thesis, keyPoints]
+conditions: {}
+graph:
+  $START:
+    - role: "analyst"
+      condition: null
+      prompt: "Analyze the topic in the task and produce a structured summary with key points."
+  analyst:
+    - role: "$END"
+      condition: null
+      prompt: "Analysis complete. Finish the workflow."

examples/debate.yaml

@@ -0,0 +1,77 @@
+name: "debate"
+description: "Structured debate between two sides. Tests cross-process session resume."
+roles:
+  against:
+    description: "Argues against the proposition"
+    goal: |
+      You are a skilled debater arguing AGAINST the proposition.
+      Be logical, cite evidence, and directly address your opponent's points.
+      Keep each argument concise (under 200 words).
+    capabilities:
+      - argumentation
+      - critical-thinking
+    procedure: |
+      1. If this is the opening, present your strongest argument against the proposition.
+      2. If responding to the other side, directly counter their points with evidence and logic.
+      3. If you find yourself genuinely convinced by the other side, you may concede.
+    output: |
+      Provide your argument in the frontmatter.
+      Set conceded to true ONLY if you are genuinely convinced and wish to stop debating.
+    frontmatter:
+      type: object
+      properties:
+        argument:
+          type: string
+        conceded:
+          type: boolean
+      required: [argument, conceded]
+  for:
+    description: "Argues for the proposition"
+    goal: |
+      You are a skilled debater arguing FOR the proposition.
+      Be logical, cite evidence, and directly address your opponent's points.
+      Keep each argument concise (under 200 words).
+    capabilities:
+      - argumentation
+      - critical-thinking
+    procedure: |
+      1. Read the opposing side's latest argument carefully.
+      2. Counter their points with evidence and logic.
+      3. If you find yourself genuinely convinced by the other side, you may concede.
+    output: |
+      Provide your argument in the frontmatter.
+      Set conceded to true ONLY if you are genuinely convinced and wish to stop debating.
+    frontmatter:
+      type: object
+      properties:
+        argument:
+          type: string
+        conceded:
+          type: boolean
+      required: [argument, conceded]
+conditions:
+  againstConceded:
+    description: "The against side conceded"
+    expression: "$last('against').conceded = true"
+  forConceded:
+    description: "The for side conceded"
+    expression: "$last('for').conceded = true"
+graph:
+  $START:
+    - role: "against"
+      condition: null
+      prompt: "Present your opening argument against the proposition."
+  against:
+    - role: "$END"
+      condition: "againstConceded"
+      prompt: "The against side conceded. Debate over."
+    - role: "for"
+      condition: null
+      prompt: "Counter the opposing argument. Address their points directly."
+  for:
+    - role: "$END"
+      condition: "forConceded"
+      prompt: "The for side conceded. Debate over."
+    - role: "against"
+      condition: null
+      prompt: "Counter the opposing argument. Address their points directly."

examples/solve-issue.yaml

@@ -3,21 +3,50 @@ description: "End-to-end issue resolution"
 roles:
   planner:
     description: "Creates implementation plan"
-    systemPrompt: "You are a planning agent. Analyze the issue and create a step-by-step plan."
-    outputSchema:
+    goal: "You are a planning agent. You analyze issues and create implementation plans grounded in the actual codebase."
+    capabilities:
+      - issue-analysis
+      - planning
+      - file-read
+      - shell
+    procedure: |
+      1. Locate the code repository:
+         - Check if the current working directory is the repo (look for package.json, .git, etc.)
+         - If the task mentions a repo URL, clone it first.
+         - If this is a new project, create the repo and note the path.
+      2. Explore the codebase — read the relevant source files mentioned in the issue. Understand the current architecture, types, and conventions (check CLAUDE.md, CONTRIBUTING.md, .cursor/rules/).
+      3. Identify which files need changes and what the changes should be, with specific code references.
+      4. Output the plan with:
+         - `repoPath`: absolute path to the repository root
+         - `plan`: detailed implementation plan with file paths and code references
+         - `steps`: concrete action items for the developer
+    output: |
+      Provide repoPath, plan summary, and steps in the frontmatter.
+      The plan MUST reference actual file paths and code structures you found by reading the source.
+      Do NOT guess — if you haven't read a file, read it before referencing it.
+    frontmatter:
       type: object
       properties:
+        repoPath:
+          type: string
         plan:
           type: string
-        steps:
-          type: array
-          items:
-            type: string
-      required: [plan, steps]
+      required: [repoPath, plan]
   developer:
     description: "Implements code changes"
-    systemPrompt: "You are a developer agent. Implement the plan."
-    outputSchema:
+    goal: "You are a developer agent. You implement code changes according to plans."
+    capabilities:
+      - file-edit
+      - shell
+      - testing
+    procedure: |
+      1. Read the planner's output to get the repoPath and implementation plan.
+      2. cd to the repoPath before making any changes.
+      3. Create a feature branch from the default branch.
+      4. Implement the plan — write code, tests, and ensure existing tests pass.
+      5. Commit your changes with a descriptive message referencing the issue.
+    output: "List all files changed and provide a summary of the implementation."
+    frontmatter:
       type: object
       properties:
         filesChanged:
@@ -29,8 +58,13 @@ roles:
       required: [filesChanged, summary]
   reviewer:
     description: "Reviews code changes"
-    systemPrompt: "You are a code reviewer. Review the implementation."
-    outputSchema:
+    goal: "You are a code reviewer. You review implementations for correctness and quality."
+    capabilities:
+      - code-review
+      - static-analysis
+    procedure: "Review the implementation against the plan. Check for bugs, edge cases, and style."
+    output: "Approve or reject with detailed comments explaining your decision."
+    frontmatter:
       type: object
       properties:
         approved:
@@ -41,19 +75,24 @@ roles:
 conditions:
   notApproved:
     description: "Reviewer rejected the implementation"
-    expression: "steps[-1].output.approved = false"
+    expression: "$last('reviewer').approved = false"
 graph:
   $START:
     - role: "planner"
       condition: null
+      prompt: "Analyze the issue described in the task and produce a detailed implementation plan."
   planner:
     - role: "developer"
       condition: null
+      prompt: "Implement the plan from the planner. Write code, tests, and ensure existing tests pass."
   developer:
     - role: "reviewer"
       condition: null
+      prompt: "Review the developer's implementation against the plan for correctness and quality."
   reviewer:
     - role: "developer"
       condition: "notApproved"
+      prompt: "The reviewer rejected your implementation. Read their feedback and fix the issues."
     - role: "$END"
       condition: null
+      prompt: "The review passed. Complete the workflow."

legacy-packages/workflow-agent-docx-diff/__tests__/agent.test.ts

@@ -1,6 +1,6 @@
 import { describe, expect, test } from "bun:test";
-import { packageDescriptor } from "../src/package-descriptor.js";
 import { createDocxDiffAgent } from "../src/agent.js";
+import { packageDescriptor } from "../src/package-descriptor.js";
 
 describe("createDocxDiffAgent", () => {
   test("returns an AdapterFn (function)", () => {

legacy-packages/workflow-agent-docx-diff/__tests__/runner.test.ts

@@ -1,8 +1,8 @@
-import { mkdirSync, writeFileSync } from "node:fs";
-import { join } from "node:path";
-import { tmpdir } from "node:os";
 import { describe, expect, mock, test } from "bun:test";
-import { ok, err } from "@uncaged/workflow-util";
+import { mkdirSync, writeFileSync } from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import { err, ok } from "@uncaged/workflow-util";
 import type { SpawnCliConfig } from "@uncaged/workflow-util-agent";
 import { runDocxDiff } from "../src/runner.js";
 
@@ -74,7 +74,12 @@ describe("runDocxDiff", () => {
   test("exit 2: throws error", async () => {
     const dir = tempDir();
     const spawnFn = makeSpawn(
-      err({ kind: "non_zero_exit", exitCode: 2, stdout: "", stderr: "fatal error" }) as MockSpawnResult,
+      err({
+        kind: "non_zero_exit",
+        exitCode: 2,
+        stdout: "",
+        stderr: "fatal error",
+      }) as MockSpawnResult,
     );
 
     await expect(

legacy-packages/workflow-agent-docx-diff/package.json

@@ -1,7 +1,11 @@
 {
   "name": "@uncaged/workflow-agent-docx-diff",
   "version": "0.1.0",
-  "files": ["src", "dist", "package.json"],
+  "files": [
+    "src",
+    "dist",
+    "package.json"
+  ],
   "type": "module",
   "types": "src/index.ts",
   "exports": {

legacy-packages/workflow-agent-docx-diff/src/agent.ts

@@ -1,7 +1,12 @@
-import * as z from "zod/v4";
 import { dirname, join } from "node:path";
-import type { AdapterFn, RoleResult, ThreadContext, WorkflowRuntime } from "@uncaged/workflow-runtime";
+import type {
+  AdapterFn,
+  RoleResult,
+  ThreadContext,
+  WorkflowRuntime,
+} from "@uncaged/workflow-runtime";
 import type { WriterMeta } from "@uncaged/workflow-template-document";
+import type * as z from "zod/v4";
 import { runDocxDiff } from "./runner.js";
 import type { DocxDiffAgentConfig } from "./types.js";
 
@@ -12,16 +17,10 @@ export function createDocxDiffAgent(config: DocxDiffAgentConfig): AdapterFn {
       if (writerStep === undefined) throw new Error("differ: no writer step found");
 
       const writerMeta = writerStep.meta as WriterMeta;
-      if (writerMeta.mode !== "edit")
-        throw new Error("differ: writer did not run in edit mode");
+      if (writerMeta.mode !== "edit") throw new Error("differ: writer did not run in edit mode");
 
       const diffDocx = join(dirname(writerMeta.outputDocx), "diff.docx");
-      const raw = await runDocxDiff(
-        config,
-        writerMeta.sourceDocx,
-        writerMeta.outputDocx,
-        diffDocx,
-      );
+      const raw = await runDocxDiff(config, writerMeta.sourceDocx, writerMeta.outputDocx, diffDocx);
 
       const meta = schema.parse(JSON.parse(raw)) as T;
       return { meta, childThread: null };

legacy-packages/workflow-agent-docx-diff/src/runner.ts

@@ -1,6 +1,6 @@
 import { stat } from "node:fs/promises";
-import { spawnCli } from "@uncaged/workflow-util-agent";
 import type { SpawnCliError } from "@uncaged/workflow-util-agent";
+import { spawnCli } from "@uncaged/workflow-util-agent";
 import type { DocxDiffAgentConfig } from "./types.js";
 
 type SpawnCliFn = typeof spawnCli;
@@ -8,8 +8,7 @@ type SpawnCliFn = typeof spawnCli;
 function throwSpawnError(e: SpawnCliError): never {
   if (e.kind === "non_zero_exit")
     throw new Error(`docx-diff failed (exit ${e.exitCode}): ${e.stderr}`);
-  if (e.kind === "timeout")
-    throw new Error("docx-diff: timed out");
+  if (e.kind === "timeout") throw new Error("docx-diff: timed out");
   throw new Error(`docx-diff: spawn failed: ${e.message}`);
 }
 

legacy-packages/workflow-agent-office/__tests__/agent.test.ts

@@ -1,6 +1,6 @@
 import { describe, expect, test } from "bun:test";
-import { packageDescriptor } from "../src/package-descriptor.js";
 import { createOfficeAgent } from "../src/agent.js";
+import { packageDescriptor } from "../src/package-descriptor.js";
 
 describe("createOfficeAgent", () => {
   test("returns an AdapterFn (function)", () => {

legacy-packages/workflow-agent-office/__tests__/runner.test.ts

@@ -1,8 +1,8 @@
-import { mkdirSync, writeFileSync } from "node:fs";
-import { join } from "node:path";
-import { tmpdir } from "node:os";
 import { describe, expect, mock, test } from "bun:test";
-import { ok, err } from "@uncaged/workflow-util";
+import { mkdirSync, writeFileSync } from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import { err, ok } from "@uncaged/workflow-util";
 import type { SpawnCliConfig } from "@uncaged/workflow-util-agent";
 import { editDocument, generateDocument } from "../src/runner.js";
 
@@ -123,7 +123,13 @@ describe("editDocument", () => {
     );
 
     await expect(
-      editDocument({ outputDir: base, command: null, timeout: null }, "te2", "edit", inputFile, spawnFn),
+      editDocument(
+        { outputDir: base, command: null, timeout: null },
+        "te2",
+        "edit",
+        inputFile,
+        spawnFn,
+      ),
     ).rejects.toThrow("spawn failed");
   });
 });

legacy-packages/workflow-agent-office/package.json

@@ -1,7 +1,11 @@
 {
   "name": "@uncaged/workflow-agent-office",
   "version": "0.1.0",
-  "files": ["src", "dist", "package.json"],
+  "files": [
+    "src",
+    "dist",
+    "package.json"
+  ],
   "type": "module",
   "types": "src/index.ts",
   "exports": {

legacy-packages/workflow-agent-office/src/agent.ts

@@ -1,6 +1,11 @@
-import * as z from "zod/v4";
-import type { AdapterFn, RoleResult, ThreadContext, WorkflowRuntime } from "@uncaged/workflow-runtime";
+import type {
+  AdapterFn,
+  RoleResult,
+  ThreadContext,
+  WorkflowRuntime,
+} from "@uncaged/workflow-runtime";
 import { createLogger } from "@uncaged/workflow-util";
+import type * as z from "zod/v4";
 import { editDocument, generateDocument } from "./runner.js";
 import type { OfficeAgentConfig } from "./types.js";
 
@@ -27,7 +32,10 @@ export function createOfficeAgent(config: OfficeAgentConfig): AdapterFn {
   return <T>(_systemPrompt: string, schema: z.ZodType<T>) =>
     async (ctx: ThreadContext, _runtime: WorkflowRuntime): Promise<RoleResult<T>> => {
       const { prompt, inputDocx } = parseStartInput(ctx.start.content);
-      log("8FQKP3NV", `office-agent: mode=${inputDocx === null ? "generate" : "edit"} thread=${ctx.threadId}`);
+      log(
+        "8FQKP3NV",
+        `office-agent: mode=${inputDocx === null ? "generate" : "edit"} thread=${ctx.threadId}`,
+      );
 
       let raw: string;
       if (inputDocx === null) {
@@ -35,7 +43,11 @@ export function createOfficeAgent(config: OfficeAgentConfig): AdapterFn {
         raw = JSON.stringify({ mode: "generate", outputDocx: result.outputDocx, sourceDocx: null });
       } else {
         const result = await editDocument(config, ctx.threadId, prompt, inputDocx);
-        raw = JSON.stringify({ mode: "edit", outputDocx: result.outputDocx, sourceDocx: result.sourceDocx });
+        raw = JSON.stringify({
+          mode: "edit",
+          outputDocx: result.outputDocx,
+          sourceDocx: result.sourceDocx,
+        });
       }
 
       const meta = schema.parse(JSON.parse(raw)) as T;

legacy-packages/workflow-agent-office/src/runner.ts

@@ -1,7 +1,7 @@
 import { copyFile, mkdir, stat } from "node:fs/promises";
 import { join } from "node:path";
-import { spawnCli } from "@uncaged/workflow-util-agent";
 import type { SpawnCliError } from "@uncaged/workflow-util-agent";
+import { spawnCli } from "@uncaged/workflow-util-agent";
 import type { OfficeAgentConfig } from "./types.js";
 
 type SpawnCliFn = typeof spawnCli;
@@ -9,8 +9,7 @@ type SpawnCliFn = typeof spawnCli;
 function throwSpawnError(e: SpawnCliError): never {
   if (e.kind === "non_zero_exit")
     throw new Error(`office-agent failed (exit ${e.exitCode}): ${e.stderr}`);
-  if (e.kind === "timeout")
-    throw new Error("office-agent: timed out");
+  if (e.kind === "timeout") throw new Error("office-agent: timed out");
   throw new Error(`office-agent: spawn failed: ${e.message}`);
 }
 

legacy-packages/workflow-dashboard/index.html

@@ -5,7 +5,7 @@
     <meta name="viewport" content="width=device-width, initial-scale=1.0" />
     <title>Workflow Dashboard</title>
     <script>
-      (function () {
+      (() => {
         var t = localStorage.getItem("theme");
         if (t === "dark" || (!t && matchMedia("(prefers-color-scheme: dark)").matches)) {
           document.documentElement.classList.add("dark");

legacy-packages/workflow-dashboard/plugins/vite-limit-line-plugin.ts

@@ -54,10 +54,14 @@ type CallExpression = {
   arguments: Array<AstExpression>;
 };
 
-type AstExpression = Identifier | MemberExpression | CallExpression | {
-  type: string;
-  [key: string]: unknown;
-};
+type AstExpression =
+  | Identifier
+  | MemberExpression
+  | CallExpression
+  | {
+      type: string;
+      [key: string]: unknown;
+    };
 
 type VariableDeclarator = {
   id: Identifier | null;
@@ -258,15 +262,21 @@ function createLimitResolver(options: LimitLineOptions): (id: string) => Resolve
 }
 
 function shouldProcess(id: string, options: LimitLineOptions): boolean {
-  return options.include.test(id) && !id.includes("node_modules") && (options.exclude === null || !options.exclude.test(id));
+  return (
+    options.include.test(id) &&
+    !id.includes("node_modules") &&
+    (options.exclude === null || !options.exclude.test(id))
+  );
 }
 
 // --- Plugin ---
 
-function viteLimitLinePlugin(
-  userOptions: Partial<LimitLineOptions> = {},
-): Array<Plugin> {
-  const options: LimitLineOptions = { ...DEFAULT_OPTIONS, ...userOptions, overrides: userOptions.overrides ?? [] };
+function viteLimitLinePlugin(userOptions: Partial<LimitLineOptions> = {}): Array<Plugin> {
+  const options: LimitLineOptions = {
+    ...DEFAULT_OPTIONS,
+    ...userOptions,
+    overrides: userOptions.overrides ?? [],
+  };
   const resolve = createLimitResolver(options);
 
   const rawCodeCache = new Map<string, string>();
@@ -358,5 +368,5 @@ function viteLimitLinePlugin(
   ];
 }
 
-export { viteLimitLinePlugin };
 export type { LimitLineOptions, LimitLineOverride };
+export { viteLimitLinePlugin };

legacy-packages/workflow-dashboard/src/components/ui/resizable-panel.tsx

@@ -55,10 +55,7 @@ export function ResizablePanel({
   }, []);
 
   return (
-    <div
-      className={cn("relative shrink-0", className)}
-      style={{ ...style, width }}
-    >
+    <div className={cn("relative shrink-0", className)} style={{ ...style, width }}>
       {children}
       <div
         className="absolute top-0 -right-1 w-2 h-full cursor-col-resize z-10 group"

legacy-packages/workflow-template-document/__tests__/document-template.test.ts

@@ -9,9 +9,7 @@ import type { DocumentMeta } from "../src/roles.js";
 
 const documentModerator = tableToModerator(documentTable);
 
-function makeCtx(
-  steps: ModeratorContext<DocumentMeta>["steps"],
-): ModeratorContext<DocumentMeta> {
+function makeCtx(steps: ModeratorContext<DocumentMeta>["steps"]): ModeratorContext<DocumentMeta> {
   return {
     threadId: "01TEST000000000000000000TR",
     depth: 0,
@@ -25,7 +23,11 @@ function writerGenerateStep(): RoleStep<DocumentMeta> {
   return {
     role: "writer",
     contentHash: "STUBHASHWRITER001",
-    meta: { mode: "generate", outputDocx: "/out/output.docx", sourceDocx: null } satisfies WriterMeta,
+    meta: {
+      mode: "generate",
+      outputDocx: "/out/output.docx",
+      sourceDocx: null,
+    } satisfies WriterMeta,
     refs: [],
     timestamp: 1,
   };
@@ -35,7 +37,11 @@ function writerEditStep(): RoleStep<DocumentMeta> {
   return {
     role: "writer",
     contentHash: "STUBHASHWRITER002",
-    meta: { mode: "edit", outputDocx: "/out/modified.docx", sourceDocx: "/out/original.docx" } satisfies WriterMeta,
+    meta: {
+      mode: "edit",
+      outputDocx: "/out/modified.docx",
+      sourceDocx: "/out/original.docx",
+    } satisfies WriterMeta,
     refs: [],
     timestamp: 1,
   };

legacy-packages/workflow-template-document/package.json

@@ -1,7 +1,11 @@
 {
   "name": "@uncaged/workflow-template-document",
   "version": "0.1.0",
-  "files": ["src", "dist", "package.json"],
+  "files": [
+    "src",
+    "dist",
+    "package.json"
+  ],
   "type": "module",
   "types": "src/index.ts",
   "exports": {

package.json

@@ -9,16 +9,18 @@
     "check": "bunx tsc --build && biome check . && bash scripts/lint-log-tags.sh",
     "typecheck": "bunx tsc --build",
     "format": "biome format --write .",
-    "test": "bun run --filter '*' test",
+    "test": "bun run --filter './packages/*' test",
     "changeset": "bunx changeset",
     "version": "bunx changeset version",
-    "release": "bun run build && bun test && npx changeset publish --no-git-tag"
+    "release": "bun run build && bun test && node scripts/publish-all.mjs"
   },
   "devDependencies": {
+    "@agentclientprotocol/sdk": "^0.22.1",
     "@biomejs/biome": "^2.4.14",
     "@changesets/cli": "^2.31.0",
     "@types/node": "^25.7.0",
     "@types/xxhashjs": "^0.2.4",
+    "@uncaged/workflow-agent-hermes": "workspace:*",
     "bun-types": "^1.3.13"
   }
 }

packages/cli-workflow/README.md

@@ -0,0 +1,128 @@
+# @uncaged/cli-workflow
+
+`uwf` CLI — thread lifecycle, workflow registry, CAS inspection, and setup.
+
+## Overview
+
+Layer 4 entry point for the workflow engine. The `uwf` binary orchestrates one step per invocation: load thread head from `threads.yaml`, run the moderator, spawn the configured agent CLI, run extract, append a CAS step node, and update the head pointer (or archive when `$END`).
+
+This package has no library `src/index.ts` — it is consumed as a CLI binary only.
+
+**Dependencies:** `@uncaged/json-cas`, `@uncaged/json-cas-fs`, `@uncaged/workflow-agent-kit`, `@uncaged/workflow-moderator`, `@uncaged/workflow-protocol`, `@uncaged/workflow-util`, `commander`, `dotenv`, `yaml`
+
+## Installation
+
+Included as the `uwf` binary when you install `@uncaged/cli-workflow`:
+
+```bash
+bun add -g @uncaged/cli-workflow
+# or from the monorepo:
+bun link packages/cli-workflow
+```
+
+## CLI Usage
+
+### Global options
+
+```
+-V, --version          Show version
+--format <json|yaml>   Output format (default: json)
+-h, --help             Show help
+```
+
+### Thread
+
+| Command | Description |
+|---------|-------------|
+| `uwf thread start <workflow> -p <prompt>` | Create a thread without executing |
+| `uwf thread step <thread-id> [--agent <cmd>] [-c <count>]` | Execute one or more moderator→agent→extract cycles |
+| `uwf thread show <thread-id>` | Show thread head pointer |
+| `uwf thread list [--all]` | List active threads (`--all` includes archived) |
+| `uwf thread steps <thread-id>` | List all steps chronologically |
+| `uwf thread read <thread-id> [--quota N] [--before <hash>] [--start]` | Render thread as readable markdown |
+| `uwf thread fork <step-hash>` | Fork from a specific step |
+| `uwf thread step-details <step-hash>` | Dump full detail node as YAML |
+| `uwf thread kill <thread-id>` | Terminate and archive |
+
+Examples:
+
+```bash
+uwf thread start solve-issue -p "Fix the login redirect bug"
+uwf thread step 01ARZ3NDEKTSV4RRFFQ69G5FAV
+uwf thread step 01ARZ3NDEKTSV4RRFFQ69G5FAV -c 3 --agent uwf-builtin
+uwf thread read 01ARZ3NDEKTSV4RRFFQ69G5FAV --quota 8000
+```
+
+### Workflow
+
+| Command | Description |
+|---------|-------------|
+| `uwf workflow put <file.yaml>` | Register a workflow from YAML |
+| `uwf workflow show <name-or-hash>` | Show workflow definition |
+| `uwf workflow list` | List registered workflows |
+
+### CAS
+
+| Command | Description |
+|---------|-------------|
+| `uwf cas get <hash> [--timestamp]` | Read a CAS node |
+| `uwf cas put <type-hash> <data>` | Store a node, print hash |
+| `uwf cas put-text <text>` | Store plain text, print hash |
+| `uwf cas has <hash>` | Check existence |
+| `uwf cas refs <hash>` | List direct references |
+| `uwf cas walk <hash>` | Recursive traversal |
+| `uwf cas reindex` | Rebuild type index |
+| `uwf cas schema list` | List registered schemas |
+| `uwf cas schema get <hash>` | Show a schema |
+
+### Setup
+
+```bash
+uwf setup
+uwf setup --provider openai --base-url https://api.openai.com/v1 \
+  --api-key sk-... --model gpt-4o --agent hermes
+```
+
+Config: `~/.uncaged/workflow/config.yaml`. API keys: `~/.uncaged/workflow/.env`.
+
+### Skill
+
+| Command | Description |
+|---------|-------------|
+| `uwf skill cli` | Print markdown reference of all uwf commands (for agent skills) |
+
+### Log
+
+| Command | Description |
+|---------|-------------|
+| `uwf log list` | List log files with sizes |
+| `uwf log show [--thread <id>] [--process <pid>] [--date YYYY-MM-DD]` | Show filtered log entries |
+| `uwf log clean [--before YYYY-MM-DD]` | Delete old log files |
+
+## Internal Structure
+
+```
+src/
+├── cli.ts              Commander entrypoint, command registration
+├── format.ts           JSON/YAML output formatting
+├── store.ts            CAS store + registry initialization
+├── validate.ts         Workflow YAML validation
+├── schemas.ts          CLI-local schema registration
+└── commands/
+    ├── thread.ts       Thread lifecycle and step execution
+    ├── workflow.ts     Workflow registry (put/show/list)
+    ├── cas.ts          CAS inspection and schema ops
+    ├── setup.ts        Interactive/non-interactive setup
+    ├── skill.ts        Built-in skill references
+    └── log.ts          Process debug log management
+```
+
+## Configuration
+
+| File | Purpose |
+|------|---------|
+| `~/.uncaged/workflow/config.yaml` | Providers, models, default agent |
+| `~/.uncaged/workflow/.env` | API keys (referenced by `apiKeyEnv` in config) |
+| `~/.uncaged/workflow/registry.yaml` | Workflow name → CAS hash |
+| `~/.uncaged/workflow/threads.yaml` | Active thread head pointers |
+| `~/.uncaged/workflow/cas/` | Content-addressed node storage |

packages/cli-workflow/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@uncaged/cli-workflow",
-  "version": "0.1.0",
+  "version": "0.5.0",
   "files": [
     "src",
     "dist",
@@ -11,8 +11,8 @@
     "uwf": "./src/cli.ts"
   },
   "dependencies": {
-    "@uncaged/json-cas": "^0.3.0",
-    "@uncaged/json-cas-fs": "^0.3.0",
+    "@uncaged/json-cas": "^0.4.0",
+    "@uncaged/json-cas-fs": "^0.4.0",
     "@uncaged/workflow-agent-kit": "workspace:^",
     "@uncaged/workflow-moderator": "workspace:^",
     "@uncaged/workflow-protocol": "workspace:^",

packages/cli-workflow/src/__tests__/log.test.ts

@@ -0,0 +1,181 @@
+import { mkdir, readdir, rm, writeFile } from "node:fs/promises";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import { afterEach, beforeEach, describe, expect, test } from "vitest";
+import { cmdLogClean, cmdLogList, cmdLogShow } from "../commands/log.js";
+
+let storageRoot: string;
+
+beforeEach(async () => {
+  storageRoot = join(tmpdir(), `uwf-log-test-${Date.now()}-${Math.random().toString(36).slice(2)}`);
+  await mkdir(join(storageRoot, "logs"), { recursive: true });
+});
+
+afterEach(async () => {
+  await rm(storageRoot, { recursive: true, force: true });
+});
+
+const entry1 = JSON.stringify({
+  ts: "2026-05-20T10:00:00.000Z",
+  pid: "1716200000000-1234",
+  tag: "W9F3RK2M",
+  msg: "process start",
+  thread: "01J1234ABCDEF",
+  workflow: "solve-issue",
+});
+
+const entry2 = JSON.stringify({
+  ts: "2026-05-20T10:00:01.000Z",
+  pid: "1716200000000-1234",
+  tag: "ABC12345",
+  msg: "step executed",
+  thread: "01J1234ABCDEF",
+  workflow: "solve-issue",
+});
+
+const entry3 = JSON.stringify({
+  ts: "2026-05-20T10:00:02.000Z",
+  pid: "1716200000000-5678",
+  tag: "XYZ98765",
+  msg: "different process",
+  thread: "01JOTHER000000",
+  workflow: "review-code",
+});
+
+const oldEntry = JSON.stringify({
+  ts: "2026-05-19T08:00:00.000Z",
+  pid: "1716200000000-9999",
+  tag: "OLD1TAG1",
+  msg: "old entry",
+  thread: "01JOLD0000000",
+  workflow: "solve-issue",
+});
+
+const olderEntry = JSON.stringify({
+  ts: "2026-05-18T08:00:00.000Z",
+  pid: "1716200000000-0001",
+  tag: "OLD2TAG2",
+  msg: "older entry",
+  thread: "01JOLDER00000",
+  workflow: "review-code",
+});
+
+async function writeLogFiles(): Promise<void> {
+  const logsDir = join(storageRoot, "logs");
+  await writeFile(join(logsDir, "2026-05-20.jsonl"), `${[entry1, entry2, entry3].join("\n")}\n`);
+  await writeFile(join(logsDir, "2026-05-19.jsonl"), `${oldEntry}\n`);
+  await writeFile(join(logsDir, "2026-05-18.jsonl"), `${olderEntry}\n`);
+}
+
+describe("cmdLogList", () => {
+  test("lists log files with sizes sorted by date descending", async () => {
+    await writeLogFiles();
+    const result = await cmdLogList(storageRoot);
+    expect(result).toHaveLength(3);
+    expect(result[0].name).toBe("2026-05-20.jsonl");
+    expect(result[0].date).toBe("2026-05-20");
+    expect(result[0].size).toBeGreaterThan(0);
+    expect(result[1].name).toBe("2026-05-19.jsonl");
+    expect(result[2].name).toBe("2026-05-18.jsonl");
+  });
+
+  test("returns empty array when no log files exist", async () => {
+    const result = await cmdLogList(storageRoot);
+    expect(result).toEqual([]);
+  });
+
+  test("returns empty array when logs directory does not exist", async () => {
+    const noLogsRoot = join(storageRoot, "nonexistent");
+    await mkdir(noLogsRoot, { recursive: true });
+    const result = await cmdLogList(noLogsRoot);
+    expect(result).toEqual([]);
+  });
+});
+
+describe("cmdLogShow", () => {
+  test("filters by thread ID", async () => {
+    await writeLogFiles();
+    const result = await cmdLogShow(storageRoot, {
+      thread: "01J1234ABCDEF",
+      process: null,
+      date: null,
+    });
+    expect(result).toHaveLength(2);
+    expect(result.every((e) => e.thread === "01J1234ABCDEF")).toBe(true);
+  });
+
+  test("filters by process ID", async () => {
+    await writeLogFiles();
+    const result = await cmdLogShow(storageRoot, {
+      thread: null,
+      process: "1716200000000-1234",
+      date: null,
+    });
+    expect(result).toHaveLength(2);
+    expect(result.every((e) => e.pid === "1716200000000-1234")).toBe(true);
+  });
+
+  test("filters by date", async () => {
+    await writeLogFiles();
+    const result = await cmdLogShow(storageRoot, {
+      thread: null,
+      process: null,
+      date: "2026-05-19",
+    });
+    expect(result).toHaveLength(1);
+    expect(result[0].msg).toBe("old entry");
+  });
+
+  test("reads all files when no date filter", async () => {
+    await writeLogFiles();
+    const result = await cmdLogShow(storageRoot, { thread: null, process: null, date: null });
+    expect(result).toHaveLength(5);
+    // sorted by ts ascending
+    expect(result[0].ts).toBe("2026-05-18T08:00:00.000Z");
+    expect(result[4].ts).toBe("2026-05-20T10:00:02.000Z");
+  });
+
+  test("returns empty when no matches", async () => {
+    await writeLogFiles();
+    const result = await cmdLogShow(storageRoot, {
+      thread: "NONEXISTENT",
+      process: null,
+      date: null,
+    });
+    expect(result).toEqual([]);
+  });
+
+  test("combined thread + date filter", async () => {
+    await writeLogFiles();
+    const result = await cmdLogShow(storageRoot, {
+      thread: "01J1234ABCDEF",
+      process: null,
+      date: "2026-05-20",
+    });
+    expect(result).toHaveLength(2);
+    expect(result.every((e) => e.thread === "01J1234ABCDEF")).toBe(true);
+  });
+});
+
+describe("cmdLogClean", () => {
+  test("deletes files before given date", async () => {
+    await writeLogFiles();
+    const result = await cmdLogClean(storageRoot, "2026-05-20");
+    expect(result.deleted).toBe(2);
+    const remaining = await readdir(join(storageRoot, "logs"));
+    expect(remaining).toEqual(["2026-05-20.jsonl"]);
+  });
+
+  test("deletes nothing when all files are newer", async () => {
+    await writeLogFiles();
+    const result = await cmdLogClean(storageRoot, "2026-05-18");
+    expect(result.deleted).toBe(0);
+  });
+
+  test("handles missing logs directory gracefully", async () => {
+    const noLogsRoot = join(storageRoot, "nonexistent");
+    await mkdir(noLogsRoot, { recursive: true });
+    const result = await cmdLogClean(noLogsRoot, "2026-05-20");
+    expect(result).toEqual({ deleted: 0 });
+  });
+});

packages/cli-workflow/src/__tests__/setup-complexity.test.ts

@@ -0,0 +1,381 @@
+import { mkdirSync, writeFileSync } from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import { afterEach, describe, expect, test, vi } from "vitest";
+import {
+  _discoverAgents,
+  _isBackspace,
+  _isTerminator,
+  _parseWhichOutput,
+  _printModelMenu,
+  _printProviderMenu,
+  _printValidationResult,
+  _resolveModelChoice,
+  _resolveProviderChoice,
+  _searchPathDirs,
+} from "../commands/setup.js";
+
+// ──────────────────────────────────────────────────────────────────────────────
+// 1a. _searchPathDirs
+// ──────────────────────────────────────────────────────────────────────────────
+
+describe("_searchPathDirs", () => {
+  test("returns empty array for empty PATH", async () => {
+    const result = await _searchPathDirs("");
+    expect(result).toEqual([]);
+  });
+
+  test("finds uwf-hermes in a single dir", async () => {
+    const dir = mkdirSync(join(tmpdir(), `uwf-test-${Date.now()}`), { recursive: true }) as
+      | string
+      | undefined;
+    const actualDir = dir ?? join(tmpdir(), `uwf-test-${Date.now()}`);
+    mkdirSync(actualDir, { recursive: true });
+    const filePath = join(actualDir, "uwf-hermes");
+    writeFileSync(filePath, "#!/bin/sh\n", { mode: 0o755 });
+    const result = await _searchPathDirs(actualDir);
+    expect(result).toContain("uwf-hermes");
+  });
+
+  test("skips non-uwf- prefixed binaries", async () => {
+    const dir = join(tmpdir(), `uwf-test-${Date.now()}-2`);
+    mkdirSync(dir, { recursive: true });
+    writeFileSync(join(dir, "hermes"), "#!/bin/sh\n", { mode: 0o755 });
+    writeFileSync(join(dir, "uwf-hermes"), "#!/bin/sh\n", { mode: 0o755 });
+    const result = await _searchPathDirs(dir);
+    expect(result).toEqual(["uwf-hermes"]);
+  });
+
+  test("skips entry named exactly 'uwf'", async () => {
+    const dir = join(tmpdir(), `uwf-test-${Date.now()}-3`);
+    mkdirSync(dir, { recursive: true });
+    writeFileSync(join(dir, "uwf"), "#!/bin/sh\n", { mode: 0o755 });
+    writeFileSync(join(dir, "uwf-hermes"), "#!/bin/sh\n", { mode: 0o755 });
+    const result = await _searchPathDirs(dir);
+    expect(result).toEqual(["uwf-hermes"]);
+  });
+
+  test("skips non-executable files", async () => {
+    const dir = join(tmpdir(), `uwf-test-${Date.now()}-4`);
+    mkdirSync(dir, { recursive: true });
+    writeFileSync(join(dir, "uwf-foo"), "#!/bin/sh\n", { mode: 0o644 });
+    const result = await _searchPathDirs(dir);
+    expect(result).toEqual([]);
+  });
+
+  test("deduplicates across PATH dirs", async () => {
+    const dir1 = join(tmpdir(), `uwf-test-${Date.now()}-5a`);
+    const dir2 = join(tmpdir(), `uwf-test-${Date.now()}-5b`);
+    mkdirSync(dir1, { recursive: true });
+    mkdirSync(dir2, { recursive: true });
+    writeFileSync(join(dir1, "uwf-hermes"), "#!/bin/sh\n", { mode: 0o755 });
+    writeFileSync(join(dir2, "uwf-hermes"), "#!/bin/sh\n", { mode: 0o755 });
+    const result = await _searchPathDirs(`${dir1}:${dir2}`);
+    expect(result).toEqual(["uwf-hermes"]);
+  });
+
+  test("returns sorted array", async () => {
+    const dir = join(tmpdir(), `uwf-test-${Date.now()}-6`);
+    mkdirSync(dir, { recursive: true });
+    writeFileSync(join(dir, "uwf-zoo"), "#!/bin/sh\n", { mode: 0o755 });
+    writeFileSync(join(dir, "uwf-alpha"), "#!/bin/sh\n", { mode: 0o755 });
+    writeFileSync(join(dir, "uwf-mid"), "#!/bin/sh\n", { mode: 0o755 });
+    const result = await _searchPathDirs(dir);
+    expect(result).toEqual(["uwf-alpha", "uwf-mid", "uwf-zoo"]);
+  });
+
+  test("skips inaccessible/nonexistent directories silently", async () => {
+    const result = await _searchPathDirs("/nonexistent-dir-xyz-abc-12345");
+    expect(result).toEqual([]);
+  });
+});
+
+// ──────────────────────────────────────────────────────────────────────────────
+// 1b. _parseWhichOutput
+// ──────────────────────────────────────────────────────────────────────────────
+
+describe("_parseWhichOutput", () => {
+  test("returns empty array for empty string", () => {
+    expect(_parseWhichOutput("")).toEqual([]);
+  });
+
+  test("parses single path", () => {
+    expect(_parseWhichOutput("/usr/local/bin/uwf-hermes")).toEqual(["uwf-hermes"]);
+  });
+
+  test("parses multiple paths", () => {
+    expect(_parseWhichOutput("/usr/local/bin/uwf-hermes\n/usr/bin/uwf-claude-code")).toEqual([
+      "uwf-claude-code",
+      "uwf-hermes",
+    ]);
+  });
+
+  test("deduplicates identical basenames from different dirs", () => {
+    expect(_parseWhichOutput("/a/uwf-hermes\n/b/uwf-hermes")).toEqual(["uwf-hermes"]);
+  });
+
+  test("skips blank lines", () => {
+    expect(_parseWhichOutput("/a/uwf-hermes\n\n/b/uwf-cursor")).toEqual([
+      "uwf-cursor",
+      "uwf-hermes",
+    ]);
+  });
+
+  test("skips entry named exactly 'uwf'", () => {
+    expect(_parseWhichOutput("/usr/bin/uwf")).toEqual([]);
+  });
+
+  test("skips basenames not starting with uwf-", () => {
+    expect(_parseWhichOutput("/usr/bin/node")).toEqual([]);
+  });
+
+  test("returns sorted array", () => {
+    expect(_parseWhichOutput("/a/uwf-zoo\n/a/uwf-alpha")).toEqual(["uwf-alpha", "uwf-zoo"]);
+  });
+});
+
+// ──────────────────────────────────────────────────────────────────────────────
+// 2a. _isTerminator
+// ──────────────────────────────────────────────────────────────────────────────
+
+describe("_isTerminator", () => {
+  test("\\n is a terminator", () => {
+    expect(_isTerminator("\n")).toBe(true);
+  });
+  test("\\r is a terminator", () => {
+    expect(_isTerminator("\r")).toBe(true);
+  });
+  test("\\u0004 (EOT) is a terminator", () => {
+    expect(_isTerminator("")).toBe(true);
+  });
+  test("regular char is not a terminator", () => {
+    expect(_isTerminator("a")).toBe(false);
+  });
+  test("empty string is not a terminator", () => {
+    expect(_isTerminator("")).toBe(false);
+  });
+});
+
+// ──────────────────────────────────────────────────────────────────────────────
+// 2b. _isBackspace
+// ──────────────────────────────────────────────────────────────────────────────
+
+describe("_isBackspace", () => {
+  test("\\u007F is a backspace", () => {
+    expect(_isBackspace("")).toBe(true);
+  });
+  test("\\b is a backspace", () => {
+    expect(_isBackspace("\b")).toBe(true);
+  });
+  test("regular char is not a backspace", () => {
+    expect(_isBackspace("x")).toBe(false);
+  });
+});
+
+// ──────────────────────────────────────────────────────────────────────────────
+// 3a. _printProviderMenu
+// ──────────────────────────────────────────────────────────────────────────────
+
+describe("_printProviderMenu", () => {
+  afterEach(() => {
+    vi.restoreAllMocks();
+  });
+
+  const providers = [
+    { name: "openai", label: "OpenAI", baseUrl: "https://api.openai.com/v1" },
+    { name: "xai", label: "xAI", baseUrl: "https://api.x.ai/v1" },
+  ] as const;
+
+  test("prints correct number of lines (one per provider + custom)", () => {
+    const lines: string[] = [];
+    vi.spyOn(console, "log").mockImplementation((msg: string) => {
+      lines.push(msg);
+    });
+    _printProviderMenu(providers);
+    // 2 providers + 1 custom = 3 lines
+    expect(lines.length).toBe(3);
+  });
+
+  test("custom option number = providers.length + 1", () => {
+    const lines: string[] = [];
+    vi.spyOn(console, "log").mockImplementation((msg: string) => {
+      lines.push(msg);
+    });
+    _printProviderMenu(providers);
+    const lastLine = lines[lines.length - 1] ?? "";
+    expect(lastLine).toMatch(/3\)/);
+  });
+
+  test("each provider line contains its label and baseUrl", () => {
+    const lines: string[] = [];
+    vi.spyOn(console, "log").mockImplementation((msg: string) => {
+      lines.push(msg);
+    });
+    _printProviderMenu(providers);
+    expect(lines[0]).toContain("OpenAI");
+    expect(lines[0]).toContain("https://api.openai.com/v1");
+    expect(lines[1]).toContain("xAI");
+    expect(lines[1]).toContain("https://api.x.ai/v1");
+  });
+});
+
+// ──────────────────────────────────────────────────────────────────────────────
+// 3b. _resolveProviderChoice
+// ──────────────────────────────────────────────────────────────────────────────
+
+describe("_resolveProviderChoice", () => {
+  const providers = [
+    { name: "openai", label: "OpenAI", baseUrl: "https://api.openai.com/v1" },
+    { name: "xai", label: "xAI", baseUrl: "https://api.x.ai/v1" },
+    { name: "deepseek", label: "DeepSeek", baseUrl: "https://api.deepseek.com/v1" },
+  ] as const;
+
+  test("valid index 1 returns first provider", () => {
+    const result = _resolveProviderChoice("1", providers);
+    expect(result).toEqual({ providerName: "openai", baseUrl: "https://api.openai.com/v1" });
+  });
+
+  test("valid index N (last preset) returns last provider", () => {
+    const result = _resolveProviderChoice("3", providers);
+    expect(result).toEqual({ providerName: "deepseek", baseUrl: "https://api.deepseek.com/v1" });
+  });
+
+  test("index providers.length+1 (custom) returns null", () => {
+    const result = _resolveProviderChoice("4", providers);
+    expect(result).toBeNull();
+  });
+
+  test("non-numeric string returns null", () => {
+    expect(_resolveProviderChoice("abc", providers)).toBeNull();
+  });
+
+  test("0 returns null (out of range)", () => {
+    expect(_resolveProviderChoice("0", providers)).toBeNull();
+  });
+
+  test("N+2 returns null (out of range)", () => {
+    expect(_resolveProviderChoice("5", providers)).toBeNull();
+  });
+
+  test("negative number returns null", () => {
+    expect(_resolveProviderChoice("-1", providers)).toBeNull();
+  });
+});
+
+// ──────────────────────────────────────────────────────────────────────────────
+// 3c. _resolveModelChoice
+// ──────────────────────────────────────────────────────────────────────────────
+
+describe("_resolveModelChoice", () => {
+  test("numeric input within range returns model at that index", () => {
+    expect(_resolveModelChoice("2", ["a", "b", "c"])).toBe("b");
+  });
+
+  test("numeric input out of range returns input as-is", () => {
+    expect(_resolveModelChoice("5", ["a"])).toBe("5");
+  });
+
+  test("non-numeric input returns input as-is", () => {
+    expect(_resolveModelChoice("gpt-4o", ["a", "b"])).toBe("gpt-4o");
+  });
+
+  test("numeric input 1 returns first model", () => {
+    expect(_resolveModelChoice("1", ["alpha", "beta"])).toBe("alpha");
+  });
+
+  test("empty models list with numeric input returns input as-is", () => {
+    expect(_resolveModelChoice("1", [])).toBe("1");
+  });
+});
+
+// ──────────────────────────────────────────────────────────────────────────────
+// 3d. _printModelMenu
+// ──────────────────────────────────────────────────────────────────────────────
+
+describe("_printModelMenu", () => {
+  afterEach(() => {
+    vi.restoreAllMocks();
+  });
+
+  test("prints all models — each model name appears in output", () => {
+    const output: string[] = [];
+    vi.spyOn(console, "log").mockImplementation((msg: string) => {
+      output.push(msg);
+    });
+    const models = ["model-a", "model-b", "model-c"];
+    _printModelMenu(models, 100);
+    const combined = output.join("\n");
+    for (const m of models) {
+      expect(combined).toContain(m);
+    }
+  });
+
+  test("single column when termCols is very small", () => {
+    const output: string[] = [];
+    vi.spyOn(console, "log").mockImplementation((msg: string) => {
+      output.push(msg);
+    });
+    _printModelMenu(["a", "b", "c"], 1);
+    // Each model on its own row → 3 lines
+    expect(output.length).toBe(3);
+  });
+
+  test("wide terminal fits multiple columns", () => {
+    const output: string[] = [];
+    vi.spyOn(console, "log").mockImplementation((msg: string) => {
+      output.push(msg);
+    });
+    const models = Array.from({ length: 6 }, (_, i) => `m${i}`);
+    _printModelMenu(models, 200);
+    // With wide terminal and short names, should fit in fewer than 6 rows
+    expect(output.length).toBeLessThan(6);
+  });
+});
+
+// ──────────────────────────────────────────────────────────────────────────────
+// 3e. _printValidationResult
+// ──────────────────────────────────────────────────────────────────────────────
+
+describe("_printValidationResult", () => {
+  afterEach(() => {
+    vi.restoreAllMocks();
+  });
+
+  test("ok=true prints success message containing '✓'", () => {
+    const lines: string[] = [];
+    vi.spyOn(console, "log").mockImplementation((msg: string) => {
+      lines.push(msg);
+    });
+    _printValidationResult({ ok: true, error: null });
+    expect(lines.join("\n")).toContain("✓");
+  });
+
+  test("ok=false prints warning message containing '⚠'", () => {
+    const lines: string[] = [];
+    vi.spyOn(console, "log").mockImplementation((msg: string) => {
+      lines.push(msg);
+    });
+    _printValidationResult({ ok: false, error: "HTTP 401" });
+    expect(lines.join("\n")).toContain("⚠");
+  });
+
+  test("ok=false includes the error string in output", () => {
+    const lines: string[] = [];
+    vi.spyOn(console, "log").mockImplementation((msg: string) => {
+      lines.push(msg);
+    });
+    _printValidationResult({ ok: false, error: "HTTP 401" });
+    expect(lines.join("\n")).toContain("HTTP 401");
+  });
+});
+
+// ──────────────────────────────────────────────────────────────────────────────
+// 4. Regression
+// ──────────────────────────────────────────────────────────────────────────────
+
+describe("_discoverAgents regression", () => {
+  test("returns an array (may be empty) — never throws", async () => {
+    const result = await _discoverAgents();
+    expect(Array.isArray(result)).toBe(true);
+  });
+});

packages/cli-workflow/src/__tests__/setup-validate.test.ts

@@ -0,0 +1,150 @@
+import { mkdtemp, rm } from "node:fs/promises";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import { afterEach, beforeEach, describe, expect, test, vi } from "vitest";
+import { cmdSetup, validateModel } from "../commands/setup.js";
+
+describe("validateModel", () => {
+  const BASE_URL = "https://api.example.com/v1";
+  const API_KEY = "sk-test-key";
+  const MODEL = "test-model";
+
+  afterEach(() => {
+    vi.restoreAllMocks();
+  });
+
+  test("success path — returns ok on 200", async () => {
+    const mockFetch = vi
+      .spyOn(globalThis, "fetch")
+      .mockResolvedValue(new Response(JSON.stringify({}), { status: 200 }));
+
+    const result = await validateModel(BASE_URL, API_KEY, MODEL);
+
+    expect(result).toEqual({ ok: true, value: undefined });
+    expect(mockFetch).toHaveBeenCalledOnce();
+
+    const [url, opts] = mockFetch.mock.calls[0]!;
+    expect(url).toBe(`${BASE_URL}/chat/completions`);
+    expect((opts as RequestInit).headers).toEqual(
+      expect.objectContaining({ Authorization: `Bearer ${API_KEY}` }),
+    );
+    const body = JSON.parse((opts as RequestInit).body as string);
+    expect(body).toEqual({
+      model: MODEL,
+      messages: [{ role: "user", content: "hi" }],
+      max_tokens: 1,
+    });
+  });
+
+  test("HTTP 401 — returns error containing 401", async () => {
+    vi.spyOn(globalThis, "fetch").mockResolvedValue(
+      new Response("Unauthorized", { status: 401, statusText: "Unauthorized" }),
+    );
+
+    const result = await validateModel(BASE_URL, API_KEY, MODEL);
+
+    expect(result.ok).toBe(false);
+    if (!result.ok) {
+      expect(result.error).toContain("401");
+    }
+  });
+
+  test("HTTP 404 — returns error containing 404", async () => {
+    vi.spyOn(globalThis, "fetch").mockResolvedValue(
+      new Response("Not Found", { status: 404, statusText: "Not Found" }),
+    );
+
+    const result = await validateModel(BASE_URL, API_KEY, MODEL);
+
+    expect(result.ok).toBe(false);
+    if (!result.ok) {
+      expect(result.error).toContain("404");
+    }
+  });
+
+  test("network timeout — returns error mentioning timeout", async () => {
+    const err = new DOMException("signal timed out", "AbortError");
+    vi.spyOn(globalThis, "fetch").mockRejectedValue(err);
+
+    const result = await validateModel(BASE_URL, API_KEY, MODEL);
+
+    expect(result.ok).toBe(false);
+    if (!result.ok) {
+      expect(result.error.toLowerCase()).toMatch(/timeout|timed out/);
+    }
+  });
+
+  test("network error (DNS/connection) — returns error mentioning connectivity", async () => {
+    vi.spyOn(globalThis, "fetch").mockRejectedValue(new TypeError("fetch failed"));
+
+    const result = await validateModel(BASE_URL, API_KEY, MODEL);
+
+    expect(result.ok).toBe(false);
+    if (!result.ok) {
+      expect(result.error.toLowerCase()).toMatch(/connect|reach|network/);
+    }
+  });
+
+  test("request body correctness", async () => {
+    const mockFetch = vi
+      .spyOn(globalThis, "fetch")
+      .mockResolvedValue(new Response(JSON.stringify({}), { status: 200 }));
+
+    await validateModel(BASE_URL, API_KEY, "my-special-model");
+
+    const body = JSON.parse((mockFetch.mock.calls[0]![1] as RequestInit).body as string);
+    expect(body).toEqual({
+      model: "my-special-model",
+      messages: [{ role: "user", content: "hi" }],
+      max_tokens: 1,
+    });
+  });
+});
+
+describe("cmdSetup with validation", () => {
+  let storageRoot: string;
+
+  beforeEach(async () => {
+    storageRoot = await mkdtemp(join(tmpdir(), "uwf-setup-validate-"));
+  });
+
+  afterEach(async () => {
+    vi.restoreAllMocks();
+    await rm(storageRoot, { recursive: true, force: true });
+  });
+
+  const setupArgs = () => ({
+    provider: "testprovider",
+    baseUrl: "https://api.test.com/v1",
+    apiKey: "sk-test",
+    model: "test-model",
+    storageRoot,
+  });
+
+  test("includes validation result on success", async () => {
+    vi.spyOn(globalThis, "fetch").mockResolvedValue(
+      new Response(JSON.stringify({}), { status: 200 }),
+    );
+
+    const result = await cmdSetup(setupArgs());
+
+    expect(result.validation).toEqual({ ok: true, value: undefined });
+    // Config files should still be written
+    expect(result.configPath).toBeTruthy();
+    expect(result.envPath).toBeTruthy();
+  });
+
+  test("includes validation failure — config still saved", async () => {
+    vi.spyOn(globalThis, "fetch").mockResolvedValue(
+      new Response("Unauthorized", { status: 401, statusText: "Unauthorized" }),
+    );
+
+    const result = await cmdSetup(setupArgs());
+
+    expect(result.validation).toBeDefined();
+    expect((result.validation as { ok: boolean }).ok).toBe(false);
+    // Config files should still be written despite validation failure
+    expect(result.configPath).toBeTruthy();
+    expect(result.envPath).toBeTruthy();
+  });
+});

packages/cli-workflow/src/__tests__/thread-read-xml-tags.test.ts

@@ -0,0 +1,683 @@
+import { mkdir, mkdtemp, rm } from "node:fs/promises";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import { bootstrap, putSchema } from "@uncaged/json-cas";
+import { createFsStore } from "@uncaged/json-cas-fs";
+import type { CasRef, ThreadId } from "@uncaged/workflow-protocol";
+import { afterEach, beforeEach, describe, expect, test } from "vitest";
+import { cmdThreadRead, THREAD_READ_DEFAULT_QUOTA } from "../commands/thread.js";
+import { registerUwfSchemas } from "../schemas.js";
+import type { UwfStore } from "../store.js";
+import { saveThreadsIndex } from "../store.js";
+
+// ── schemas used in tests ────────────────────────────────────────────────────
+
+const TURN_SCHEMA = {
+  title: "hermes-turn",
+  type: "object" as const,
+  required: ["index", "role", "content"],
+  properties: {
+    index: { type: "integer" as const },
+    role: { type: "string" as const },
+    content: { type: "string" as const },
+    toolCalls: {
+      anyOf: [
+        { type: "array" as const, items: { type: "object" as const } },
+        { type: "null" as const },
+      ],
+    },
+    reasoning: { anyOf: [{ type: "string" as const }, { type: "null" as const }] },
+  },
+  additionalProperties: false,
+};
+
+const DETAIL_SCHEMA = {
+  title: "hermes-detail",
+  type: "object" as const,
+  required: ["sessionId", "model", "duration", "turnCount", "turns"],
+  properties: {
+    sessionId: { type: "string" as const },
+    model: { type: "string" as const },
+    duration: { type: "integer" as const },
+    turnCount: { type: "integer" as const },
+    turns: {
+      type: "array" as const,
+      items: { type: "string" as const, format: "cas_ref" },
+    },
+  },
+  additionalProperties: false,
+};
+
+// ── helpers ───────────────────────────────────────────────────────────────────
+
+async function makeUwfStore(storageRoot: string): Promise<UwfStore> {
+  const casDir = join(storageRoot, "cas");
+  await mkdir(casDir, { recursive: true });
+  const store = createFsStore(casDir);
+  const schemas = await registerUwfSchemas(store);
+  return { storageRoot, store, schemas };
+}
+
+async function registerDetailSchemas(store: ReturnType<typeof createFsStore>) {
+  await bootstrap(store);
+  const [turn, detail] = await Promise.all([
+    putSchema(store, TURN_SCHEMA),
+    putSchema(store, DETAIL_SCHEMA),
+  ]);
+  return { turn, detail };
+}
+
+// ── fixture ───────────────────────────────────────────────────────────────────
+
+let tmpDir: string;
+
+beforeEach(async () => {
+  tmpDir = await mkdtemp(join(tmpdir(), "cli-uwf-test-"));
+});
+
+afterEach(async () => {
+  await rm(tmpDir, { recursive: true, force: true });
+});
+
+// ── thread read XML tag isolation ─────────────────────────────────────────────
+
+describe("thread read XML tag isolation", () => {
+  test("scenario 1: wraps output in XML tags instead of heading", async () => {
+    const uwf = await makeUwfStore(tmpDir);
+    const detailSchemas = await registerDetailSchemas(uwf.store);
+
+    const workflowHash = await uwf.store.put(uwf.schemas.workflow, {
+      name: "test-wf",
+      description: "desc",
+      roles: {
+        planner: {
+          description: "Planner",
+          goal: "You are a planning agent. Your task is to...",
+          capabilities: [],
+          procedure: "Plan the work.",
+          output: "Summarize the plan.",
+          meta: "placeholder00" as CasRef,
+        },
+      },
+      conditions: {},
+      graph: {},
+    });
+
+    const startHash = await uwf.store.put(uwf.schemas.startNode, {
+      workflow: workflowHash,
+      prompt: "Fix issue #459",
+    });
+
+    const outputHash = await uwf.store.put(uwf.schemas.workflow, {
+      name: "out",
+      description: "",
+      roles: {},
+      conditions: {},
+      graph: {},
+    });
+
+    const turnHash = await uwf.store.put(detailSchemas.turn, {
+      index: 0,
+      role: "assistant",
+      content:
+        "---\nstatus: ready\nplan: CMWGHQKT58RY4\n---\n\n# Analysis Complete\n## Issue Summary\nThe issue requires XML tag isolation.",
+      toolCalls: null,
+      reasoning: null,
+    });
+    const detailHash = await uwf.store.put(detailSchemas.detail, {
+      sessionId: "sx",
+      model: "mx",
+      duration: 500,
+      turnCount: 1,
+      turns: [turnHash],
+    });
+
+    const stepHash = await uwf.store.put(uwf.schemas.stepNode, {
+      start: startHash,
+      prev: null,
+      role: "planner",
+      output: outputHash,
+      detail: detailHash,
+      agent: "uwf-claude-code",
+    });
+
+    const threadId = "01JTEST0000000000000001" as ThreadId;
+    await saveThreadsIndex(tmpDir, { [threadId]: stepHash });
+
+    const markdown = await cmdThreadRead(tmpDir, threadId, THREAD_READ_DEFAULT_QUOTA, null, false);
+
+    // Should wrap output in XML tags
+    expect(markdown).toContain("<output>");
+    expect(markdown).toContain("</output>");
+
+    // Should not have ### Content heading
+    expect(markdown).not.toContain("### Content");
+
+    // Should preserve markdown headings inside output tags
+    expect(markdown).toContain("# Analysis Complete");
+    expect(markdown).toContain("## Issue Summary");
+  });
+
+  test("scenario 2: wraps prompt in XML tags", async () => {
+    const uwf = await makeUwfStore(tmpDir);
+    const detailSchemas = await registerDetailSchemas(uwf.store);
+
+    const workflowHash = await uwf.store.put(uwf.schemas.workflow, {
+      name: "test-wf",
+      description: "desc",
+      roles: {
+        planner: {
+          description: "Planner",
+          goal: "You are a planning agent. Your task is to analyze and plan.",
+          capabilities: [],
+          procedure: "Plan the work.",
+          output: "Summarize the plan.",
+          meta: "placeholder00" as CasRef,
+        },
+      },
+      conditions: {},
+      graph: {},
+    });
+
+    const startHash = await uwf.store.put(uwf.schemas.startNode, {
+      workflow: workflowHash,
+      prompt: "Fix issue",
+    });
+
+    const outputHash = await uwf.store.put(uwf.schemas.workflow, {
+      name: "out",
+      description: "",
+      roles: {},
+      conditions: {},
+      graph: {},
+    });
+
+    const turnHash = await uwf.store.put(detailSchemas.turn, {
+      index: 0,
+      role: "assistant",
+      content: "---\nstatus: ready\n---\n\nContent here...",
+      toolCalls: null,
+      reasoning: null,
+    });
+    const detailHash = await uwf.store.put(detailSchemas.detail, {
+      sessionId: "sx",
+      model: "mx",
+      duration: 500,
+      turnCount: 1,
+      turns: [turnHash],
+    });
+
+    const stepHash = await uwf.store.put(uwf.schemas.stepNode, {
+      start: startHash,
+      prev: null,
+      role: "planner",
+      output: outputHash,
+      detail: detailHash,
+      agent: "uwf-claude-code",
+    });
+
+    const threadId = "01JTEST0000000000000002" as ThreadId;
+    await saveThreadsIndex(tmpDir, { [threadId]: stepHash });
+
+    const markdown = await cmdThreadRead(tmpDir, threadId, THREAD_READ_DEFAULT_QUOTA, null, false);
+
+    // Should wrap prompt in XML tags
+    expect(markdown).toContain("<prompt>");
+    expect(markdown).toContain("</prompt>");
+    expect(markdown).toContain("You are a planning agent. Your task is to analyze and plan.");
+
+    // Should not have ### Prompt heading
+    expect(markdown).not.toContain("### Prompt");
+
+    // Should wrap output in XML tags
+    expect(markdown).toContain("<output>");
+    expect(markdown).toContain("</output>");
+  });
+
+  test("scenario 3: same role repeated does not show prompt twice", async () => {
+    const uwf = await makeUwfStore(tmpDir);
+
+    const workflowHash = await uwf.store.put(uwf.schemas.workflow, {
+      name: "test-wf",
+      description: "desc",
+      roles: {
+        writer: {
+          description: "Writer",
+          goal: "You are a writer agent.",
+          capabilities: [],
+          procedure: "Write content.",
+          output: "Summarize writing.",
+          meta: "placeholder00" as CasRef,
+        },
+      },
+      conditions: {},
+      graph: {},
+    });
+
+    const startHash = await uwf.store.put(uwf.schemas.startNode, {
+      workflow: workflowHash,
+      prompt: "Write something",
+    });
+
+    const outputHash = await uwf.store.put(uwf.schemas.workflow, {
+      name: "out",
+      description: "",
+      roles: {},
+      conditions: {},
+      graph: {},
+    });
+
+    const step1 = await uwf.store.put(uwf.schemas.stepNode, {
+      start: startHash,
+      prev: null,
+      role: "writer",
+      output: outputHash,
+      detail: null,
+      agent: "uwf-test",
+    });
+
+    const step2 = await uwf.store.put(uwf.schemas.stepNode, {
+      start: startHash,
+      prev: step1 as CasRef,
+      role: "writer",
+      output: outputHash,
+      detail: null,
+      agent: "uwf-test",
+    });
+
+    const threadId = "01JTEST0000000000000003" as ThreadId;
+    await saveThreadsIndex(tmpDir, { [threadId]: step2 });
+
+    const markdown = await cmdThreadRead(tmpDir, threadId, THREAD_READ_DEFAULT_QUOTA, null, false);
+
+    // Should only show prompt tags once
+    const promptCount = (markdown.match(/<prompt>/g) ?? []).length;
+    expect(promptCount).toBe(1);
+  });
+
+  test("scenario 4: step with no detail shows no output tags", async () => {
+    const uwf = await makeUwfStore(tmpDir);
+
+    const workflowHash = await uwf.store.put(uwf.schemas.workflow, {
+      name: "test-wf",
+      description: "desc",
+      roles: {
+        worker: {
+          description: "Worker",
+          goal: "You are a worker agent.",
+          capabilities: [],
+          procedure: "Do work.",
+          output: "Summarize work.",
+          meta: "placeholder00" as CasRef,
+        },
+      },
+      conditions: {},
+      graph: {},
+    });
+
+    const startHash = await uwf.store.put(uwf.schemas.startNode, {
+      workflow: workflowHash,
+      prompt: "Do stuff",
+    });
+
+    const outputHash = await uwf.store.put(uwf.schemas.workflow, {
+      name: "out",
+      description: "",
+      roles: {},
+      conditions: {},
+      graph: {},
+    });
+
+    const stepHash = await uwf.store.put(uwf.schemas.stepNode, {
+      start: startHash,
+      prev: null,
+      role: "worker",
+      output: outputHash,
+      detail: null,
+      agent: "uwf-test",
+    });
+
+    const threadId = "01JTEST0000000000000004" as ThreadId;
+    await saveThreadsIndex(tmpDir, { [threadId]: stepHash });
+
+    const markdown = await cmdThreadRead(tmpDir, threadId, THREAD_READ_DEFAULT_QUOTA, null, false);
+
+    // Should not have output tags
+    expect(markdown).not.toContain("<output>");
+    expect(markdown).not.toContain("</output>");
+
+    // Step header should still be displayed
+    expect(markdown).toContain("## Step 1: worker");
+
+    // Prompt should still be shown
+    expect(markdown).toContain("<prompt>");
+  });
+
+  test("scenario 5: empty content shows no output tags", async () => {
+    const uwf = await makeUwfStore(tmpDir);
+
+    const workflowHash = await uwf.store.put(uwf.schemas.workflow, {
+      name: "test-wf",
+      description: "desc",
+      roles: {},
+      conditions: {},
+      graph: {},
+    });
+
+    const startHash = await uwf.store.put(uwf.schemas.startNode, {
+      workflow: workflowHash,
+      prompt: "Do stuff",
+    });
+
+    const outputHash = await uwf.store.put(uwf.schemas.workflow, {
+      name: "out",
+      description: "",
+      roles: {},
+      conditions: {},
+      graph: {},
+    });
+
+    // A detail ref that doesn't exist → extractLastAssistantContent returns null
+    const missingDetailRef = "missingdetail0" as CasRef;
+
+    const stepHash = await uwf.store.put(uwf.schemas.stepNode, {
+      start: startHash,
+      prev: null,
+      role: "worker",
+      output: outputHash,
+      detail: missingDetailRef,
+      agent: "uwf-test",
+    });
+
+    const threadId = "01JTEST0000000000000005" as ThreadId;
+    await saveThreadsIndex(tmpDir, { [threadId]: stepHash });
+
+    const markdown = await cmdThreadRead(tmpDir, threadId, THREAD_READ_DEFAULT_QUOTA, null, false);
+
+    // Should not have output tags
+    expect(markdown).not.toContain("<output>");
+    expect(markdown).not.toContain("</output>");
+  });
+
+  test("scenario 6: thread read with --start flag shows task section", async () => {
+    const uwf = await makeUwfStore(tmpDir);
+
+    const workflowHash = await uwf.store.put(uwf.schemas.workflow, {
+      name: "test-wf",
+      description: "desc",
+      roles: {
+        roleA: {
+          description: "Role A",
+          goal: "Goal for roleA",
+          capabilities: [],
+          procedure: "Do stuff.",
+          output: "Output.",
+          meta: "placeholder00" as CasRef,
+        },
+      },
+      conditions: {},
+      graph: {},
+    });
+
+    const startHash = await uwf.store.put(uwf.schemas.startNode, {
+      workflow: workflowHash,
+      prompt: "Initial prompt",
+    });
+
+    const outputHash = await uwf.store.put(uwf.schemas.workflow, {
+      name: "out",
+      description: "",
+      roles: {},
+      conditions: {},
+      graph: {},
+    });
+
+    const stepHash = await uwf.store.put(uwf.schemas.stepNode, {
+      start: startHash,
+      prev: null,
+      role: "roleA",
+      output: outputHash,
+      detail: null,
+      agent: "uwf-test",
+    });
+
+    const threadId = "01JTEST0000000000000006" as ThreadId;
+    await saveThreadsIndex(tmpDir, { [threadId]: stepHash });
+
+    const markdown = await cmdThreadRead(tmpDir, threadId, THREAD_READ_DEFAULT_QUOTA, null, true);
+
+    // Should include task section
+    expect(markdown).toContain("# Thread");
+    expect(markdown).toContain("## Task");
+    expect(markdown).toContain("Initial prompt");
+
+    // Prompts should use XML tags
+    expect(markdown).toContain("<prompt>");
+  });
+
+  test("scenario 7: thread read with --before parameter", async () => {
+    const uwf = await makeUwfStore(tmpDir);
+
+    const workflowHash = await uwf.store.put(uwf.schemas.workflow, {
+      name: "test-wf",
+      description: "desc",
+      roles: {
+        roleA: {
+          description: "Role A",
+          goal: "Goal for roleA",
+          capabilities: [],
+          procedure: "Do stuff.",
+          output: "Output.",
+          meta: "placeholder00" as CasRef,
+        },
+        roleB: {
+          description: "Role B",
+          goal: "Goal for roleB",
+          capabilities: [],
+          procedure: "Do stuff.",
+          output: "Output.",
+          meta: "placeholder00" as CasRef,
+        },
+        roleC: {
+          description: "Role C",
+          goal: "Goal for roleC",
+          capabilities: [],
+          procedure: "Do stuff.",
+          output: "Output.",
+          meta: "placeholder00" as CasRef,
+        },
+      },
+      conditions: {},
+      graph: {},
+    });
+
+    const startHash = await uwf.store.put(uwf.schemas.startNode, {
+      workflow: workflowHash,
+      prompt: "Initial prompt",
+    });
+
+    const outputHash = await uwf.store.put(uwf.schemas.workflow, {
+      name: "out",
+      description: "",
+      roles: {},
+      conditions: {},
+      graph: {},
+    });
+
+    const step1 = await uwf.store.put(uwf.schemas.stepNode, {
+      start: startHash,
+      prev: null,
+      role: "roleA",
+      output: outputHash,
+      detail: null,
+      agent: "uwf-test",
+    });
+
+    const step2 = await uwf.store.put(uwf.schemas.stepNode, {
+      start: startHash,
+      prev: step1 as CasRef,
+      role: "roleB",
+      output: outputHash,
+      detail: null,
+      agent: "uwf-test",
+    });
+
+    const step3 = await uwf.store.put(uwf.schemas.stepNode, {
+      start: startHash,
+      prev: step2 as CasRef,
+      role: "roleC",
+      output: outputHash,
+      detail: null,
+      agent: "uwf-test",
+    });
+
+    const threadId = "01JTEST0000000000000007" as ThreadId;
+    await saveThreadsIndex(tmpDir, { [threadId]: step3 });
+
+    const markdown = await cmdThreadRead(
+      tmpDir,
+      threadId,
+      THREAD_READ_DEFAULT_QUOTA,
+      step2 as CasRef,
+      false,
+    );
+
+    // Should only show roleA
+    expect(markdown).toContain("roleA");
+    expect(markdown).not.toContain("roleB");
+    expect(markdown).not.toContain("roleC");
+
+    // Should use XML tags
+    expect(markdown).toContain("<prompt>");
+  });
+
+  test("scenario 9: special characters in content are preserved", async () => {
+    const uwf = await makeUwfStore(tmpDir);
+    const detailSchemas = await registerDetailSchemas(uwf.store);
+
+    const workflowHash = await uwf.store.put(uwf.schemas.workflow, {
+      name: "test-wf",
+      description: "desc",
+      roles: {
+        writer: {
+          description: "Writer",
+          goal: "You are a writer.",
+          capabilities: [],
+          procedure: "Write content.",
+          output: "Summarize.",
+          meta: "placeholder00" as CasRef,
+        },
+      },
+      conditions: {},
+      graph: {},
+    });
+
+    const startHash = await uwf.store.put(uwf.schemas.startNode, {
+      workflow: workflowHash,
+      prompt: "Write something",
+    });
+
+    const outputHash = await uwf.store.put(uwf.schemas.workflow, {
+      name: "out",
+      description: "",
+      roles: {},
+      conditions: {},
+      graph: {},
+    });
+
+    const turnHash = await uwf.store.put(detailSchemas.turn, {
+      index: 0,
+      role: "assistant",
+      content: "Content with <special> & characters > like <this>",
+      toolCalls: null,
+      reasoning: null,
+    });
+    const detailHash = await uwf.store.put(detailSchemas.detail, {
+      sessionId: "sx",
+      model: "mx",
+      duration: 500,
+      turnCount: 1,
+      turns: [turnHash],
+    });
+
+    const stepHash = await uwf.store.put(uwf.schemas.stepNode, {
+      start: startHash,
+      prev: null,
+      role: "writer",
+      output: outputHash,
+      detail: detailHash,
+      agent: "uwf-test",
+    });
+
+    const threadId = "01JTEST0000000000000008" as ThreadId;
+    await saveThreadsIndex(tmpDir, { [threadId]: stepHash });
+
+    const markdown = await cmdThreadRead(tmpDir, threadId, THREAD_READ_DEFAULT_QUOTA, null, false);
+
+    // Special characters should be preserved as-is
+    expect(markdown).toContain("Content with <special> & characters > like <this>");
+  });
+
+  test("scenario 10: quota limit with XML tags", async () => {
+    const uwf = await makeUwfStore(tmpDir);
+
+    const workflowHash = await uwf.store.put(uwf.schemas.workflow, {
+      name: "test-wf",
+      description: "desc",
+      roles: {
+        roleA: {
+          description: "Role A",
+          goal: "Goal for roleA",
+          capabilities: [],
+          procedure: "Do stuff.",
+          output: "Output.",
+          meta: "placeholder00" as CasRef,
+        },
+      },
+      conditions: {},
+      graph: {},
+    });
+
+    const startHash = await uwf.store.put(uwf.schemas.startNode, {
+      workflow: workflowHash,
+      prompt: "Initial prompt",
+    });
+
+    const outputHash = await uwf.store.put(uwf.schemas.workflow, {
+      name: "out",
+      description: "",
+      roles: {},
+      conditions: {},
+      graph: {},
+    });
+
+    const steps: CasRef[] = [];
+    let prev: CasRef | null = null;
+    for (let i = 0; i < 5; i++) {
+      const step = (await uwf.store.put(uwf.schemas.stepNode, {
+        start: startHash,
+        prev,
+        role: "roleA",
+        output: outputHash,
+        detail: null,
+        agent: "uwf-test",
+      })) as CasRef;
+      steps.push(step);
+      prev = step;
+    }
+
+    const threadId = "01JTEST0000000000000009" as ThreadId;
+    await saveThreadsIndex(tmpDir, { [threadId]: steps[steps.length - 1]! });
+
+    // Use very small quota
+    const markdown = await cmdThreadRead(tmpDir, threadId, 1, null, false);
+
+    // Should have skip hint
+    expect(markdown).toContain("earlier step");
+
+    // Should have XML tags for displayed steps
+    if (markdown.includes("<prompt>")) {
+      expect(markdown).toContain("</prompt>");
+    }
+  });
+});

packages/cli-workflow/src/__tests__/thread-step-count.test.ts

@@ -0,0 +1,71 @@
+import { execFileSync } from "node:child_process";
+import { join } from "node:path";
+import { describe, expect, test } from "vitest";
+
+const CLI_PATH = join(import.meta.dirname, "..", "cli.js");
+
+function runCli(args: string[]): { stdout: string; stderr: string; exitCode: number } {
+  try {
+    const stdout = execFileSync("bun", ["run", CLI_PATH, ...args], {
+      encoding: "utf8",
+      env: { ...process.env, WORKFLOW_STORAGE_ROOT: "/tmp/uwf-test-nonexistent" },
+      stdio: ["ignore", "pipe", "pipe"],
+    });
+    return { stdout, stderr: "", exitCode: 0 };
+  } catch (e: unknown) {
+    const err = e as NodeJS.ErrnoException & { stdout?: string; stderr?: string; status?: number };
+    return {
+      stdout: err.stdout ?? "",
+      stderr: err.stderr ?? "",
+      exitCode: err.status ?? 1,
+    };
+  }
+}
+
+describe("thread step --count CLI parsing", () => {
+  test("--help shows -c/--count option", () => {
+    const result = runCli(["thread", "step", "--help"]);
+    expect(result.stdout).toContain("--count");
+    expect(result.stdout).toContain("-c");
+  });
+
+  test("description says 'one or more steps'", () => {
+    const result = runCli(["thread", "step", "--help"]);
+    expect(result.stdout).toContain("one or more steps");
+  });
+});
+
+describe("cmdThreadStep count logic", () => {
+  test("count=0 fails with validation error", () => {
+    const result = runCli(["thread", "step", "FAKE_THREAD_ID", "-c", "0"]);
+    expect(result.exitCode).not.toBe(0);
+    expect(result.stderr).toContain("positive integer");
+  });
+
+  test("negative count fails with validation error", () => {
+    const result = runCli(["thread", "step", "FAKE_THREAD_ID", "-c", "-1"]);
+    expect(result.exitCode).not.toBe(0);
+    expect(result.stderr).toContain("positive integer");
+  });
+
+  test("non-integer count fails with validation error", () => {
+    const result = runCli(["thread", "step", "FAKE_THREAD_ID", "-c", "1.5"]);
+    expect(result.exitCode).not.toBe(0);
+    expect(result.stderr).toContain("positive integer");
+  });
+
+  test("count=1 is the default (no -c flag)", () => {
+    // Without -c, it should attempt to run 1 step (failing on missing thread, not on count validation)
+    const result = runCli(["thread", "step", "FAKE_THREAD_ID"]);
+    expect(result.exitCode).not.toBe(0);
+    // Should NOT contain "positive integer" error — should fail on thread lookup instead
+    expect(result.stderr).not.toContain("positive integer");
+  });
+
+  test("count=3 passes validation (fails on thread lookup)", () => {
+    const result = runCli(["thread", "step", "FAKE_THREAD_ID", "-c", "3"]);
+    expect(result.exitCode).not.toBe(0);
+    // Should NOT contain "positive integer" error — should fail on thread/storage lookup
+    expect(result.stderr).not.toContain("positive integer");
+  });
+});

packages/cli-workflow/src/__tests__/thread.test.ts

@@ -198,10 +198,10 @@ describe("extractLastAssistantContent", () => {
   });
 });
 
-// ── cmdThreadRead: ### Content section ───────────────────────────────────────
+// ── cmdThreadRead: <output> section ──────────────────────────────────────────
 
-describe("cmdThreadRead ### Content section", () => {
-  test("includes ### Content before ### Output when detail has assistant turns", async () => {
+describe("cmdThreadRead <output> section", () => {
+  test("includes <output> tags when detail has assistant turns", async () => {
     const uwf = await makeUwfStore(tmpDir);
     const detailSchemas = await registerDetailSchemas(uwf.store);
 
@@ -211,8 +211,11 @@ describe("cmdThreadRead ### Content section", () => {
       roles: {
         writer: {
           description: "Write",
-          systemPrompt: "You are a writer.",
-          outputSchema: "placeholder00" as CasRef,
+          goal: "You are a writer.",
+          capabilities: [],
+          procedure: "Write content as requested.",
+          output: "Summarize what was written.",
+          meta: "placeholder00" as CasRef,
         },
       },
       conditions: {},
@@ -261,17 +264,13 @@ describe("cmdThreadRead ### Content section", () => {
 
     const markdown = await cmdThreadRead(tmpDir, threadId, THREAD_READ_DEFAULT_QUOTA, null, false);
 
-    expect(markdown).toContain("### Content");
+    expect(markdown).toContain("<output>");
+    expect(markdown).toContain("</output>");
     expect(markdown).toContain("The assistant response text");
-
-    const contentIdx = markdown.indexOf("### Content");
-    const outputIdx = markdown.indexOf("### Output");
-    expect(contentIdx).toBeGreaterThanOrEqual(0);
-    expect(outputIdx).toBeGreaterThanOrEqual(0);
-    expect(contentIdx).toBeLessThan(outputIdx);
+    expect(markdown).not.toContain("### Content");
   });
 
-  test("omits ### Content when detail has no matching assistant turns", async () => {
+  test("omits <output> tags when detail has no matching assistant turns", async () => {
     const uwf = await makeUwfStore(tmpDir);
 
     const workflowHash = await uwf.store.put(uwf.schemas.workflow, {
@@ -310,8 +309,9 @@ describe("cmdThreadRead ### Content section", () => {
 
     const markdown = await cmdThreadRead(tmpDir, threadId, THREAD_READ_DEFAULT_QUOTA, null, false);
 
+    expect(markdown).not.toContain("<output>");
+    expect(markdown).not.toContain("</output>");
     expect(markdown).not.toContain("### Content");
-    expect(markdown).toContain("### Output");
   });
 });
 
@@ -384,8 +384,266 @@ describe("cmdThreadStepDetails", () => {
       content: "done",
     });
   });
+});
 
+// ── cmdThreadRead: <prompt> deduplication ────────────────────────────────────
+
+describe("cmdThreadRead <prompt> deduplication", () => {
+  async function makeThreadWithRoles(uwf: UwfStore, roles: string[]): Promise<string> {
+    const roleMap: Record<string, unknown> = {};
+    for (const r of [...new Set(roles)]) {
+      roleMap[r] = {
+        description: r,
+        goal: `Goal for ${r}`,
+        capabilities: [],
+        procedure: "Do stuff.",
+        output: "Output.",
+        meta: "placeholder00" as CasRef,
+      };
+    }
+    const workflowHash = await uwf.store.put(uwf.schemas.workflow, {
+      name: "dedup-wf",
+      description: "desc",
+      roles: roleMap,
+      conditions: {},
+      graph: {},
+    });
+    const startHash = await uwf.store.put(uwf.schemas.startNode, {
+      workflow: workflowHash,
+      prompt: "Start",
+    });
+    const outputHash = await uwf.store.put(uwf.schemas.workflow, {
+      name: "out",
+      description: "",
+      roles: {},
+      conditions: {},
+      graph: {},
+    });
+
+    let prev: string | null = null;
+    let stepHash = "";
+    for (const role of roles) {
+      stepHash = await uwf.store.put(uwf.schemas.stepNode, {
+        start: startHash,
+        prev: prev as CasRef | null,
+        role,
+        output: outputHash,
+        detail: null,
+        agent: "uwf-test",
+      });
+      prev = stepHash;
+    }
+    return stepHash;
+  }
+
+  test("same consecutive role shows <prompt> once", async () => {
+    const uwf = await makeUwfStore(tmpDir);
+    const headHash = await makeThreadWithRoles(uwf, ["writer", "writer"]);
+    const threadId = "01JTEST0000000000000003" as ThreadId;
+    await saveThreadsIndex(tmpDir, { [threadId]: headHash });
+
+    const markdown = await cmdThreadRead(tmpDir, threadId, THREAD_READ_DEFAULT_QUOTA, null, false);
+    const count = (markdown.match(/<prompt>/g) ?? []).length;
+    expect(count).toBe(1);
+  });
+
+  test("different consecutive roles each show <prompt>", async () => {
+    const uwf = await makeUwfStore(tmpDir);
+    const headHash = await makeThreadWithRoles(uwf, ["planner", "coder"]);
+    const threadId = "01JTEST0000000000000004" as ThreadId;
+    await saveThreadsIndex(tmpDir, { [threadId]: headHash });
+
+    const markdown = await cmdThreadRead(tmpDir, threadId, THREAD_READ_DEFAULT_QUOTA, null, false);
+    const count = (markdown.match(/<prompt>/g) ?? []).length;
+    expect(count).toBe(2);
+  });
+
+  test("non-consecutive same role shows <prompt> twice", async () => {
+    const uwf = await makeUwfStore(tmpDir);
+    const headHash = await makeThreadWithRoles(uwf, ["roleA", "roleB", "roleA"]);
+    const threadId = "01JTEST0000000000000005" as ThreadId;
+    await saveThreadsIndex(tmpDir, { [threadId]: headHash });
+
+    const markdown = await cmdThreadRead(tmpDir, threadId, THREAD_READ_DEFAULT_QUOTA, null, false);
+    const count = (markdown.match(/<prompt>/g) ?? []).length;
+    expect(count).toBe(2);
+  });
+});
+
+// ── cmdThreadRead: showStart / before / quota ─────────────────────────────────
+
+describe("cmdThreadRead start section / before / quota", () => {
+  async function makeSimpleThread(
+    uwf: UwfStore,
+    roles: string[],
+  ): Promise<{ startHash: CasRef; stepHashes: CasRef[] }> {
+    const uniqueRoles = [...new Set(roles)];
+    const workflowHash = await uwf.store.put(uwf.schemas.workflow, {
+      name: "simple-wf",
+      description: "desc",
+      roles: Object.fromEntries(
+        uniqueRoles.map((r) => [
+          r,
+          {
+            description: r,
+            goal: `Goal for ${r}`,
+            capabilities: [],
+            procedure: "Do stuff.",
+            output: "Output.",
+            meta: "placeholder00" as CasRef,
+          },
+        ]),
+      ),
+      conditions: {},
+      graph: {},
+    });
+    const startHash = (await uwf.store.put(uwf.schemas.startNode, {
+      workflow: workflowHash,
+      prompt: "Initial prompt",
+    })) as CasRef;
+    const outputHash = await uwf.store.put(uwf.schemas.workflow, {
+      name: "out",
+      description: "",
+      roles: {},
+      conditions: {},
+      graph: {},
+    });
+
+    const stepHashes: CasRef[] = [];
+    let prev: CasRef | null = null;
+    for (const role of roles) {
+      const stepHash = (await uwf.store.put(uwf.schemas.stepNode, {
+        start: startHash,
+        prev,
+        role,
+        output: outputHash,
+        detail: null,
+        agent: "uwf-test",
+      })) as CasRef;
+      stepHashes.push(stepHash);
+      prev = stepHash;
+    }
+    return { startHash, stepHashes };
+  }
+
+  test("showStart=true includes # Thread header and ## Task section", async () => {
+    const uwf = await makeUwfStore(tmpDir);
+    const { stepHashes } = await makeSimpleThread(uwf, ["roleA"]);
+    const threadId = "01JTEST0000000000000006" as ThreadId;
+    await saveThreadsIndex(tmpDir, { [threadId]: stepHashes[stepHashes.length - 1]! });
+
+    const markdown = await cmdThreadRead(tmpDir, threadId, THREAD_READ_DEFAULT_QUOTA, null, true);
+    expect(markdown).toContain("# Thread");
+    expect(markdown).toContain("## Task");
+    expect(markdown).toContain("Initial prompt");
+  });
+
+  test("showStart=false with before=null still shows # Thread header (default behavior)", async () => {
+    const uwf = await makeUwfStore(tmpDir);
+    const { stepHashes } = await makeSimpleThread(uwf, ["roleA"]);
+    const threadId = "01JTEST0000000000000007" as ThreadId;
+    await saveThreadsIndex(tmpDir, { [threadId]: stepHashes[stepHashes.length - 1]! });
+
+    // When before=null, the start section is always shown regardless of showStart
+    const markdown = await cmdThreadRead(tmpDir, threadId, THREAD_READ_DEFAULT_QUOTA, null, false);
+    expect(markdown).toContain("# Thread");
+    expect(markdown).toContain("## Task");
+  });
+
+  test("before filter: only steps before the given hash appear", async () => {
+    const uwf = await makeUwfStore(tmpDir);
+    const { stepHashes } = await makeSimpleThread(uwf, ["roleA", "roleB", "roleC"]);
+    const [_hashA, hashB, hashC] = stepHashes as [CasRef, CasRef, CasRef];
+    const threadId = "01JTEST0000000000000008" as ThreadId;
+    await saveThreadsIndex(tmpDir, { [threadId]: hashC });
+
+    const markdown = await cmdThreadRead(tmpDir, threadId, THREAD_READ_DEFAULT_QUOTA, hashB, false);
+    expect(markdown).toContain("roleA");
+    expect(markdown).not.toContain("roleB");
+    expect(markdown).not.toContain("roleC");
+  });
+
+  test("quota=1 limits output and includes skip hint", async () => {
+    const uwf = await makeUwfStore(tmpDir);
+    const { stepHashes } = await makeSimpleThread(uwf, ["roleA", "roleB", "roleC"]);
+    const threadId = "01JTEST000000000000000A" as ThreadId;
+    await saveThreadsIndex(tmpDir, { [threadId]: stepHashes[stepHashes.length - 1]! });
+
+    const markdown = await cmdThreadRead(tmpDir, threadId, 1, null, false);
+    expect(markdown).toContain("earlier step");
+  });
+
+  test("all steps fit in quota: no skip hint", async () => {
+    const uwf = await makeUwfStore(tmpDir);
+    const { stepHashes } = await makeSimpleThread(uwf, ["roleA"]);
+    const threadId = "01JTEST000000000000000B" as ThreadId;
+    await saveThreadsIndex(tmpDir, { [threadId]: stepHashes[0]! });
+
+    const markdown = await cmdThreadRead(tmpDir, threadId, THREAD_READ_DEFAULT_QUOTA, null, false);
+    expect(markdown).not.toContain("earlier step");
+  });
+});
+
+// ── Tests that call process.exit must be last ─────────────────────────────────
+
+describe("cmdThreadStepDetails (process.exit tests - must be last)", () => {
   test("throws when step hash does not exist", async () => {
     await expect(cmdThreadStepDetails(tmpDir, "nonexistenth0" as CasRef)).rejects.toThrow();
   });
+
+  test("before with unknown hash rejects", async () => {
+    const _uwf = await makeUwfStore(tmpDir);
+    const casDir = join(tmpDir, "cas");
+    await mkdir(casDir, { recursive: true });
+    const store = createFsStore(casDir);
+    const schemas = await registerUwfSchemas(store);
+    const uwfStore: UwfStore = { storageRoot: tmpDir, store, schemas };
+
+    const workflowHash = await uwfStore.store.put(uwfStore.schemas.workflow, {
+      name: "wf2",
+      description: "",
+      roles: {
+        roleA: {
+          description: "r",
+          goal: "g",
+          capabilities: [],
+          procedure: "p",
+          output: "o",
+          meta: "placeholder00" as CasRef,
+        },
+      },
+      conditions: {},
+      graph: {},
+    });
+    const startHash = await uwfStore.store.put(uwfStore.schemas.startNode, {
+      workflow: workflowHash,
+      prompt: "p",
+    });
+    const outputHash = await uwfStore.store.put(uwfStore.schemas.workflow, {
+      name: "out",
+      description: "",
+      roles: {},
+      conditions: {},
+      graph: {},
+    });
+    const stepHash = await uwfStore.store.put(uwfStore.schemas.stepNode, {
+      start: startHash,
+      prev: null,
+      role: "roleA",
+      output: outputHash,
+      detail: null,
+      agent: "uwf-test",
+    });
+    await saveThreadsIndex(tmpDir, { ["01JTEST000000000000000C" as ThreadId]: stepHash as CasRef });
+
+    await expect(
+      cmdThreadRead(
+        tmpDir,
+        "01JTEST000000000000000C" as ThreadId,
+        THREAD_READ_DEFAULT_QUOTA,
+        "unknownhash0" as CasRef,
+        false,
+      ),
+    ).rejects.toThrow();
+  });
 });

packages/cli-workflow/src/__tests__/workflow-resolution.test.ts

@@ -0,0 +1,367 @@
+import { mkdir, mkdtemp, rm, writeFile } from "node:fs/promises";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import { createFsStore } from "@uncaged/json-cas-fs";
+import type { CasRef, WorkflowPayload } from "@uncaged/workflow-protocol";
+import { afterEach, beforeEach, describe, expect, test } from "vitest";
+import { stringify } from "yaml";
+import { cmdThreadStart } from "../commands/thread.js";
+import { registerUwfSchemas } from "../schemas.js";
+import type { UwfStore } from "../store.js";
+import { loadWorkflowRegistry, saveWorkflowRegistry } from "../store.js";
+
+// ── helpers ───────────────────────────────────────────────────────────────────
+
+async function makeUwfStore(storageRoot: string): Promise<UwfStore> {
+  const casDir = join(storageRoot, "cas");
+  await mkdir(casDir, { recursive: true });
+  const store = createFsStore(casDir);
+  const schemas = await registerUwfSchemas(store);
+  return { storageRoot, store, schemas };
+}
+
+async function storeWorkflow(uwf: UwfStore, name: string): Promise<CasRef> {
+  const payload: WorkflowPayload = {
+    name,
+    description: "Test workflow",
+    roles: {},
+    conditions: {},
+    graph: {},
+  };
+  return await uwf.store.put(uwf.schemas.workflow, payload);
+}
+
+async function createWorkflowYaml(name: string, version: string | null = null): Promise<string> {
+  const payload: WorkflowPayload = {
+    name,
+    description: version !== null ? `Test workflow (${version})` : "Test workflow",
+    roles: {},
+    conditions: {},
+    graph: {},
+  };
+  const yaml = stringify(payload);
+  return yaml;
+}
+
+// ── fixture ───────────────────────────────────────────────────────────────────
+
+let tmpDir: string;
+let storageRoot: string;
+let projectRoot: string;
+
+beforeEach(async () => {
+  tmpDir = await mkdtemp(join(tmpdir(), "cli-uwf-wf-resolve-test-"));
+  storageRoot = join(tmpDir, "storage");
+  projectRoot = join(tmpDir, "project");
+  await mkdir(storageRoot, { recursive: true });
+  await mkdir(projectRoot, { recursive: true });
+});
+
+afterEach(async () => {
+  await rm(tmpDir, { recursive: true, force: true });
+});
+
+// ── Strategy 1: CAS Hash Resolution ───────────────────────────────────────────
+
+describe("Strategy 1: CAS Hash Resolution", () => {
+  test("should resolve valid 13-char Crockford Base32 hash", async () => {
+    const uwf = await makeUwfStore(storageRoot);
+    const hash = await storeWorkflow(uwf, "test-workflow");
+
+    const result = await cmdThreadStart(storageRoot, hash, "test prompt", projectRoot);
+
+    expect(result.workflow).toBe(hash);
+    expect(result.thread).toMatch(/^[0-9A-HJKMNP-TV-Z]{26}$/);
+  });
+
+  test("should fail on invalid hash format (non-Crockford characters)", async () => {
+    await makeUwfStore(storageRoot);
+
+    await expect(
+      cmdThreadStart(storageRoot, "123456789ABCD", "prompt", projectRoot),
+    ).rejects.toThrow();
+  });
+
+  test("should fail on valid-format hash not present in CAS", async () => {
+    await makeUwfStore(storageRoot);
+    const fakeHash = "0000000000000"; // valid format, doesn't exist
+
+    await expect(cmdThreadStart(storageRoot, fakeHash, "prompt", projectRoot)).rejects.toThrow();
+  });
+
+  test("should reject 40-char hex hash (legacy format not supported)", async () => {
+    await makeUwfStore(storageRoot);
+    const hexHash = "a".repeat(40);
+
+    await expect(cmdThreadStart(storageRoot, hexHash, "prompt", projectRoot)).rejects.toThrow();
+  });
+});
+
+// ── Strategy 2: File Path Resolution ──────────────────────────────────────────
+
+describe("Strategy 2: File Path Resolution", () => {
+  test("should load workflow from absolute file path", async () => {
+    await makeUwfStore(storageRoot);
+    const yamlPath = join(tmpDir, "test-workflow.yaml");
+    await writeFile(yamlPath, await createWorkflowYaml("test-workflow"));
+
+    const result = await cmdThreadStart(storageRoot, yamlPath, "prompt", projectRoot);
+
+    expect(result.workflow).toMatch(/^[0-9A-HJKMNP-TV-Z]{13}$/);
+    const uwf = await makeUwfStore(storageRoot);
+    const node = uwf.store.get(result.workflow);
+    expect(node).not.toBeNull();
+    if (node !== null) {
+      expect((node.payload as WorkflowPayload).name).toBe("test-workflow");
+    }
+  });
+
+  test("should load workflow from relative file path", async () => {
+    await makeUwfStore(storageRoot);
+    const yamlPath = "test-workflow.yaml";
+    await writeFile(join(projectRoot, yamlPath), await createWorkflowYaml("test-workflow"));
+
+    const result = await cmdThreadStart(storageRoot, yamlPath, "prompt", projectRoot);
+
+    expect(result.workflow).toMatch(/^[0-9A-HJKMNP-TV-Z]{13}$/);
+  });
+
+  test("should fail when file path does not exist", async () => {
+    await makeUwfStore(storageRoot);
+
+    await expect(
+      cmdThreadStart(storageRoot, "./nonexistent.yaml", "prompt", projectRoot),
+    ).rejects.toThrow();
+  });
+
+  test("should fail on invalid YAML syntax in file", async () => {
+    await makeUwfStore(storageRoot);
+    const yamlPath = join(tmpDir, "bad-syntax.yaml");
+    await writeFile(yamlPath, "invalid: yaml: : :");
+
+    await expect(cmdThreadStart(storageRoot, yamlPath, "prompt", projectRoot)).rejects.toThrow();
+  });
+
+  test("should fail on valid YAML with invalid WorkflowPayload shape", async () => {
+    await makeUwfStore(storageRoot);
+    const yamlPath = join(tmpDir, "invalid-workflow.yaml");
+    await writeFile(yamlPath, "name: test\n# missing roles, conditions, and graph");
+
+    await expect(cmdThreadStart(storageRoot, yamlPath, "prompt", projectRoot)).rejects.toThrow();
+  });
+
+  test("should enforce filename matches workflow name", async () => {
+    await makeUwfStore(storageRoot);
+    const yamlPath = join(tmpDir, "solve-issue.yaml");
+    await writeFile(yamlPath, await createWorkflowYaml("wrong-name"));
+
+    await expect(cmdThreadStart(storageRoot, yamlPath, "prompt", projectRoot)).rejects.toThrow();
+  });
+});
+
+// ── Strategy 3: Local Discovery (Parent Traversal) ────────────────────────────
+
+describe("Strategy 3: Local Discovery", () => {
+  test("should find workflow in current directory .workflow/", async () => {
+    await makeUwfStore(storageRoot);
+    const workflowDir = join(projectRoot, ".workflow");
+    await mkdir(workflowDir, { recursive: true });
+    await writeFile(join(workflowDir, "solve-issue.yaml"), await createWorkflowYaml("solve-issue"));
+
+    const result = await cmdThreadStart(storageRoot, "solve-issue", "prompt", projectRoot);
+
+    expect(result.workflow).toMatch(/^[0-9A-HJKMNP-TV-Z]{13}$/);
+    const uwf = await makeUwfStore(storageRoot);
+    const node = uwf.store.get(result.workflow);
+    expect(node).not.toBeNull();
+    if (node !== null) {
+      expect((node.payload as WorkflowPayload).name).toBe("solve-issue");
+    }
+  });
+
+  test("should find workflow in parent directory .workflow/", async () => {
+    await makeUwfStore(storageRoot);
+    const workflowDir = join(projectRoot, ".workflow");
+    await mkdir(workflowDir, { recursive: true });
+    await writeFile(join(workflowDir, "solve-issue.yaml"), await createWorkflowYaml("solve-issue"));
+
+    const subdir = join(projectRoot, "packages", "cli-workflow", "src");
+    await mkdir(subdir, { recursive: true });
+
+    const result = await cmdThreadStart(storageRoot, "solve-issue", "prompt", subdir);
+
+    expect(result.workflow).toMatch(/^[0-9A-HJKMNP-TV-Z]{13}$/);
+  });
+
+  test("should stop at filesystem root when traversing", async () => {
+    await makeUwfStore(storageRoot);
+    const deepPath = join(tmpDir, "deep", "path", "that", "does", "not", "have", "workflow");
+    await mkdir(deepPath, { recursive: true });
+
+    await expect(cmdThreadStart(storageRoot, "nonexistent", "prompt", deepPath)).rejects.toThrow();
+  });
+
+  test("should prefer .workflow/ over .workflows/ directory", async () => {
+    await makeUwfStore(storageRoot);
+    const workflowDir = join(projectRoot, ".workflow");
+    const workflowsDir = join(projectRoot, ".workflows");
+    await mkdir(workflowDir, { recursive: true });
+    await mkdir(workflowsDir, { recursive: true });
+
+    await writeFile(
+      join(workflowDir, "solve-issue.yaml"),
+      await createWorkflowYaml("solve-issue", "1"),
+    );
+    await writeFile(
+      join(workflowsDir, "solve-issue.yaml"),
+      await createWorkflowYaml("solve-issue", "2"),
+    );
+
+    const result = await cmdThreadStart(storageRoot, "solve-issue", "prompt", projectRoot);
+
+    const uwf = await makeUwfStore(storageRoot);
+    const node = uwf.store.get(result.workflow);
+    expect(node).not.toBeNull();
+    if (node !== null) {
+      expect((node.payload as WorkflowPayload).description).toBe("Test workflow (1)");
+    }
+  });
+
+  test("should support .yml extension in local discovery", async () => {
+    await makeUwfStore(storageRoot);
+    const workflowDir = join(projectRoot, ".workflow");
+    await mkdir(workflowDir, { recursive: true });
+    await writeFile(join(workflowDir, "solve-issue.yml"), await createWorkflowYaml("solve-issue"));
+
+    const result = await cmdThreadStart(storageRoot, "solve-issue", "prompt", projectRoot);
+
+    expect(result.workflow).toMatch(/^[0-9A-HJKMNP-TV-Z]{13}$/);
+  });
+});
+
+// ── Strategy 4: Global Registry Fallback ──────────────────────────────────────
+
+describe("Strategy 4: Global Registry Resolution", () => {
+  test("should resolve workflow from global registry when not found locally", async () => {
+    const uwf = await makeUwfStore(storageRoot);
+    const hash = await storeWorkflow(uwf, "deploy-pipeline");
+    const registry = await loadWorkflowRegistry(storageRoot);
+    registry["deploy-pipeline"] = hash;
+    await saveWorkflowRegistry(storageRoot, registry);
+
+    const isolatedRoot = join(tmpDir, "isolated");
+    await mkdir(isolatedRoot, { recursive: true });
+
+    const result = await cmdThreadStart(storageRoot, "deploy-pipeline", "prompt", isolatedRoot);
+
+    expect(result.workflow).toBe(hash);
+  });
+
+  test("should fail when workflow not found in any strategy", async () => {
+    await makeUwfStore(storageRoot);
+
+    await expect(cmdThreadStart(storageRoot, "nonexistent", "prompt", tmpDir)).rejects.toThrow();
+  });
+});
+
+// ── Strategy Priority Order ───────────────────────────────────────────────────
+
+describe("Resolution Priority", () => {
+  test("should use explicit file path over local discovery", async () => {
+    await makeUwfStore(storageRoot);
+
+    // Setup: Create workflow in .workflow/ AND as explicit file
+    const workflowDir = join(projectRoot, ".workflow");
+    await mkdir(workflowDir, { recursive: true });
+    await writeFile(
+      join(workflowDir, "solve-issue.yaml"),
+      await createWorkflowYaml("solve-issue", "discovery"),
+    );
+
+    const explicitPath = join(projectRoot, "custom-solve-issue.yaml");
+    await writeFile(explicitPath, await createWorkflowYaml("custom-solve-issue", "explicit"));
+
+    // Execute with explicit path
+    const result = await cmdThreadStart(storageRoot, explicitPath, "prompt", projectRoot);
+
+    const uwf = await makeUwfStore(storageRoot);
+    const node = uwf.store.get(result.workflow);
+    expect(node).not.toBeNull();
+    if (node !== null) {
+      expect((node.payload as WorkflowPayload).description).toBe("Test workflow (explicit)");
+    }
+  });
+
+  test("should use local discovery over global registry", async () => {
+    const uwf = await makeUwfStore(storageRoot);
+
+    // Setup: Register globally
+    const globalHash = await storeWorkflow(uwf, "solve-issue");
+    const registry = await loadWorkflowRegistry(storageRoot);
+    registry["solve-issue"] = globalHash;
+    await saveWorkflowRegistry(storageRoot, registry);
+
+    // Setup: Create local .workflow/
+    const workflowDir = join(projectRoot, ".workflow");
+    await mkdir(workflowDir, { recursive: true });
+    const localYaml = await createWorkflowYaml("solve-issue", "local");
+    await writeFile(join(workflowDir, "solve-issue.yaml"), localYaml);
+
+    const result = await cmdThreadStart(storageRoot, "solve-issue", "prompt", projectRoot);
+
+    const uwf2 = await makeUwfStore(storageRoot);
+    const node = uwf2.store.get(result.workflow);
+    expect(node).not.toBeNull();
+    if (node !== null) {
+      expect((node.payload as WorkflowPayload).description).toBe("Test workflow (local)");
+    }
+  });
+});
+
+// ── Edge Cases ────────────────────────────────────────────────────────────────
+
+describe("Edge Cases", () => {
+  test("should treat '13-char-string.yaml' as file path, not CAS hash", async () => {
+    await makeUwfStore(storageRoot);
+    const fileName = "0123456789ABC.yaml"; // 13 chars + .yaml
+    await writeFile(join(projectRoot, fileName), await createWorkflowYaml("0123456789ABC"));
+
+    const result = await cmdThreadStart(storageRoot, fileName, "prompt", projectRoot);
+
+    expect(result.workflow).toMatch(/^[0-9A-HJKMNP-TV-Z]{13}$/);
+  });
+
+  test("should handle workflow names containing slashes as file paths", async () => {
+    await makeUwfStore(storageRoot);
+    const filePath = "subdir/solve-issue.yaml";
+    const fullPath = join(projectRoot, filePath);
+    await mkdir(join(projectRoot, "subdir"), { recursive: true });
+    await writeFile(fullPath, await createWorkflowYaml("solve-issue"));
+
+    const result = await cmdThreadStart(storageRoot, filePath, "prompt", projectRoot);
+
+    expect(result.workflow).toMatch(/^[0-9A-HJKMNP-TV-Z]{13}$/);
+  });
+
+  test("should handle absolute paths correctly", async () => {
+    await makeUwfStore(storageRoot);
+    const absPath = join(tmpDir, "abs-workflow.yaml");
+    await writeFile(absPath, await createWorkflowYaml("abs-workflow"));
+
+    const result = await cmdThreadStart(storageRoot, absPath, "prompt", projectRoot);
+
+    expect(result.workflow).toMatch(/^[0-9A-HJKMNP-TV-Z]{13}$/);
+  });
+
+  test("should fail on empty workflow ID", async () => {
+    await makeUwfStore(storageRoot);
+
+    await expect(cmdThreadStart(storageRoot, "", "prompt", projectRoot)).rejects.toThrow();
+  });
+
+  test("should fail on whitespace-only workflow ID", async () => {
+    await makeUwfStore(storageRoot);
+
+    await expect(cmdThreadStart(storageRoot, "   ", "prompt", projectRoot)).rejects.toThrow();
+  });
+});

packages/cli-workflow/src/background/background.ts

@@ -0,0 +1,147 @@
+import { mkdir, readdir, readFile, rename, rm, writeFile } from "node:fs/promises";
+import { join } from "node:path";
+import type { RunningThreadItem, ThreadId } from "@uncaged/workflow-protocol";
+
+import type { RunningMarker } from "./types.js";
+
+/**
+ * Get the path to the running markers directory.
+ */
+export function getRunningDir(storageRoot: string): string {
+  return join(storageRoot, "running");
+}
+
+/**
+ * Get the path to a specific thread's marker file.
+ */
+export function getMarkerPath(storageRoot: string, threadId: ThreadId): string {
+  return join(getRunningDir(storageRoot), `${threadId}.json`);
+}
+
+/**
+ * Check if a PID is still running.
+ * Returns true if the process exists, false otherwise.
+ */
+export function isPidAlive(pid: number): boolean {
+  try {
+    // process.kill with signal 0 checks existence without killing
+    process.kill(pid, 0);
+    return true;
+  } catch {
+    // ESRCH means process doesn't exist
+    return false;
+  }
+}
+
+/**
+ * Create a marker file for a running thread.
+ * Writes to a temp file in the same directory, then atomically renames.
+ */
+export async function createMarker(storageRoot: string, marker: RunningMarker): Promise<void> {
+  const runningDir = getRunningDir(storageRoot);
+  await mkdir(runningDir, { recursive: true });
+
+  const markerPath = getMarkerPath(storageRoot, marker.thread);
+  const tempPath = join(runningDir, `.${marker.thread}-${process.pid}.tmp`);
+
+  const content = JSON.stringify(marker, null, 2);
+  await writeFile(tempPath, content, "utf8");
+  await rename(tempPath, markerPath);
+}
+
+/**
+ * Delete a marker file for a thread.
+ */
+export async function deleteMarker(storageRoot: string, threadId: ThreadId): Promise<void> {
+  const markerPath = getMarkerPath(storageRoot, threadId);
+  try {
+    await rm(markerPath);
+  } catch {
+    // Ignore errors if file doesn't exist
+  }
+}
+
+/**
+ * Read a marker file. Returns null if file doesn't exist or is invalid.
+ */
+export async function readMarker(
+  storageRoot: string,
+  threadId: ThreadId,
+): Promise<RunningMarker | null> {
+  const markerPath = getMarkerPath(storageRoot, threadId);
+  try {
+    const content = await readFile(markerPath, "utf8");
+    const marker = JSON.parse(content) as RunningMarker;
+    return marker;
+  } catch {
+    return null;
+  }
+}
+
+/**
+ * List all running threads, filtering out stale markers.
+ */
+export async function listRunningThreads(storageRoot: string): Promise<RunningThreadItem[]> {
+  const runningDir = getRunningDir(storageRoot);
+
+  let files: string[];
+  try {
+    files = await readdir(runningDir);
+  } catch {
+    // Directory doesn't exist or can't be read
+    return [];
+  }
+
+  const results: RunningThreadItem[] = [];
+
+  for (const filename of files) {
+    if (!filename.endsWith(".json")) {
+      continue;
+    }
+
+    const threadId = filename.slice(0, -5) as ThreadId;
+    const marker = await readMarker(storageRoot, threadId);
+
+    if (marker === null) {
+      // Invalid marker file
+      continue;
+    }
+
+    if (!isPidAlive(marker.pid)) {
+      // Stale marker - process no longer exists
+      await deleteMarker(storageRoot, threadId);
+      continue;
+    }
+
+    results.push({
+      thread: marker.thread,
+      workflow: marker.workflow,
+      pid: marker.pid,
+      startedAt: marker.startedAt,
+    });
+  }
+
+  return results;
+}
+
+/**
+ * Check if a thread is currently executing in the background.
+ * Returns the marker if running, null otherwise.
+ */
+export async function isThreadRunning(
+  storageRoot: string,
+  threadId: ThreadId,
+): Promise<RunningMarker | null> {
+  const marker = await readMarker(storageRoot, threadId);
+  if (marker === null) {
+    return null;
+  }
+
+  if (!isPidAlive(marker.pid)) {
+    // Stale marker
+    await deleteMarker(storageRoot, threadId);
+    return null;
+  }
+
+  return marker;
+}

packages/cli-workflow/src/background/index.ts

@@ -0,0 +1,11 @@
+export {
+  createMarker,
+  deleteMarker,
+  getMarkerPath,
+  getRunningDir,
+  isPidAlive,
+  isThreadRunning,
+  listRunningThreads,
+  readMarker,
+} from "./background.js";
+export type { RunningMarker } from "./types.js";

packages/cli-workflow/src/background/types.ts

@@ -0,0 +1,9 @@
+import type { CasRef, ThreadId } from "@uncaged/workflow-protocol";
+
+/** Marker file stored at ~/.uncaged/workflow/running/<thread-id>.json */
+export type RunningMarker = {
+  thread: ThreadId;
+  workflow: CasRef;
+  pid: number;
+  startedAt: number;
+};

packages/cli-workflow/src/cli.ts

@@ -7,18 +7,22 @@ import {
   cmdCasGet,
   cmdCasHas,
   cmdCasPut,
+  cmdCasPutText,
   cmdCasRefs,
   cmdCasReindex,
   cmdCasSchemaGet,
   cmdCasSchemaList,
   cmdCasWalk,
 } from "./commands/cas.js";
+import { cmdLogClean, cmdLogList, cmdLogShow } from "./commands/log.js";
 import { cmdSetup, cmdSetupInteractive } from "./commands/setup.js";
+import { cmdSkillCli } from "./commands/skill.js";
 import {
   cmdThreadFork,
   cmdThreadKill,
   cmdThreadList,
   cmdThreadRead,
+  cmdThreadRunning,
   cmdThreadShow,
   cmdThreadStart,
   cmdThreadStep,
@@ -45,7 +49,12 @@ function runAction(action: () => Promise<void>): void {
 
 const program = new Command();
 
-program.name("uwf").description("Stateless workflow CLI");
+// eslint-disable-next-line -- dynamic import for version
+const pkg = await import("../package.json", { with: { type: "json" } });
+program
+  .name("uwf")
+  .description("Stateless workflow CLI")
+  .version(pkg.default.version, "-V, --version");
 program.option("--format <fmt>", "Output format: json or yaml", "json");
 
 const workflow = program.command("workflow").description("Workflow registry and CAS");
@@ -80,7 +89,7 @@ workflow
   .action(() => {
     const storageRoot = resolveStorageRoot();
     runAction(async () => {
-      const result = await cmdWorkflowList(storageRoot);
+      const result = await cmdWorkflowList(storageRoot, process.cwd());
       writeOutput(result);
     });
   });
@@ -95,24 +104,52 @@ thread
   .action((workflow: string, opts: { prompt: string }) => {
     const storageRoot = resolveStorageRoot();
     runAction(async () => {
-      const result = await cmdThreadStart(storageRoot, workflow, opts.prompt);
+      const result = await cmdThreadStart(storageRoot, workflow, opts.prompt, process.cwd());
       writeOutput(result);
     });
   });
 
 thread
   .command("step")
-  .description("Execute one step")
+  .description("Execute one or more steps")
   .argument("<thread-id>", "Thread ULID")
   .option("--agent <cmd>", "Override agent command")
-  .action((threadId: string, opts: { agent: string | undefined }) => {
-    const storageRoot = resolveStorageRoot();
-    runAction(async () => {
-      const agentOverride = opts.agent ?? null;
-      const result = await cmdThreadStep(storageRoot, threadId, agentOverride);
-      writeOutput(result);
-    });
-  });
+  .option("-c, --count <number>", "Number of steps to run (default: 1)")
+  .option("--background", "Run in background and return immediately")
+  .option("--_background-worker", "Internal flag for background worker process", false)
+  .action(
+    (
+      threadId: string,
+      opts: {
+        agent: string | undefined;
+        count: string | undefined;
+        background: boolean;
+        _backgroundWorker: boolean;
+      },
+    ) => {
+      const storageRoot = resolveStorageRoot();
+      runAction(async () => {
+        const agentOverride = opts.agent ?? null;
+        const count = opts.count !== undefined ? Number(opts.count) : 1;
+        const background = opts.background ?? false;
+        const backgroundWorker = opts._backgroundWorker ?? false;
+
+        const results = await cmdThreadStep(
+          storageRoot,
+          threadId,
+          agentOverride,
+          count,
+          background,
+          backgroundWorker,
+        );
+        if (results.length === 1) {
+          writeOutput(results[0]);
+        } else {
+          writeOutput(results);
+        }
+      });
+    },
+  );
 
 thread
   .command("show")
@@ -138,6 +175,17 @@ thread
     });
   });
 
+thread
+  .command("running")
+  .description("List threads currently executing in the background")
+  .action(() => {
+    const storageRoot = resolveStorageRoot();
+    runAction(async () => {
+      const result = await cmdThreadRunning(storageRoot);
+      writeOutput(result);
+    });
+  });
+
 thread
   .command("kill")
   .description("Terminate and archive a thread")
@@ -215,6 +263,15 @@ thread
     });
   });
 
+const skill = program.command("skill").description("Built-in skill references for agents");
+
+skill
+  .command("cli")
+  .description("Print a markdown reference of all uwf commands")
+  .action(() => {
+    console.log(cmdSkillCli());
+  });
+
 program
   .command("setup")
   .description("Configure provider, model, and agent")
@@ -280,6 +337,17 @@ cas
     });
   });
 
+cas
+  .command("put-text")
+  .description("Store a plain text string, print its hash")
+  .argument("<text>", "Text content to store")
+  .action((text: string) => {
+    const storageRoot = resolveStorageRoot();
+    runAction(async () => {
+      writeOutput(await cmdCasPutText(storageRoot, text));
+    });
+  });
+
 cas
   .command("has")
   .description("Check if a hash exists")
@@ -346,6 +414,55 @@ casSchema
     });
   });
 
+const log = program.command("log").description("Process-level debug logs");
+
+log
+  .command("list")
+  .description("List log files with sizes")
+  .action(() => {
+    const storageRoot = resolveStorageRoot();
+    runAction(async () => {
+      const result = await cmdLogList(storageRoot);
+      writeOutput(result);
+    });
+  });
+
+log
+  .command("show")
+  .description("Show and filter log entries")
+  .option("--thread <thread-id>", "Filter by thread ID")
+  .option("--process <pid>", "Filter by process ID")
+  .option("--date <date>", "Filter by date (YYYY-MM-DD)")
+  .action(
+    (opts: {
+      thread: string | undefined;
+      process: string | undefined;
+      date: string | undefined;
+    }) => {
+      const storageRoot = resolveStorageRoot();
+      runAction(async () => {
+        const result = await cmdLogShow(storageRoot, {
+          thread: opts.thread ?? null,
+          process: opts.process ?? null,
+          date: opts.date ?? null,
+        });
+        writeOutput(result);
+      });
+    },
+  );
+
+log
+  .command("clean")
+  .description("Delete log files older than given date")
+  .requiredOption("--before <date>", "Delete files before this date (YYYY-MM-DD)")
+  .action((opts: { before: string }) => {
+    const storageRoot = resolveStorageRoot();
+    runAction(async () => {
+      const result = await cmdLogClean(storageRoot, opts.before);
+      writeOutput(result);
+    });
+  });
+
 program.parseAsync(process.argv).catch((e: unknown) => {
   const message = e instanceof Error ? e.message : String(e);
   process.stderr.write(`${message}\n`);

packages/cli-workflow/src/commands/cas.ts

@@ -1,10 +1,12 @@
 import { readFileSync } from "node:fs";
 import { join } from "node:path";
 
-import type { Hash, JSONSchema, Store } from "@uncaged/json-cas";
-import { bootstrap, getSchema, refs, walk } from "@uncaged/json-cas";
+import type { JSONSchema, Store } from "@uncaged/json-cas";
+import { bootstrap, getSchema, putSchema, refs, walk } from "@uncaged/json-cas";
 import { createFsStore } from "@uncaged/json-cas-fs";
 
+import { TEXT_SCHEMA } from "../schemas.js";
+
 // ---- Helpers ----
 
 function openStore(storageRoot: string): Store {
@@ -53,18 +55,12 @@ export async function cmdCasPut(
   return { hash };
 }
 
-export async function cmdCasHas(
-  storageRoot: string,
-  hash: string,
-): Promise<{ exists: boolean }> {
+export async function cmdCasHas(storageRoot: string, hash: string): Promise<{ exists: boolean }> {
   const store = openStore(storageRoot);
   return { exists: store.has(hash) };
 }
 
-export async function cmdCasRefs(
-  storageRoot: string,
-  hash: string,
-): Promise<{ refs: string[] }> {
+export async function cmdCasRefs(storageRoot: string, hash: string): Promise<{ refs: string[] }> {
   const store = openStore(storageRoot);
   const node = store.get(hash);
   if (node === null) {
@@ -73,10 +69,7 @@ export async function cmdCasRefs(
   return { refs: refs(store, node) };
 }
 
-export async function cmdCasWalk(
-  storageRoot: string,
-  hash: string,
-): Promise<{ hashes: string[] }> {
+export async function cmdCasWalk(storageRoot: string, hash: string): Promise<{ hashes: string[] }> {
   const store = openStore(storageRoot);
   const result: string[] = [];
   walk(store, hash, (h) => {
@@ -90,9 +83,7 @@ export type SchemaListEntry = {
   title: string;
 };
 
-export async function cmdCasSchemaList(
-  storageRoot: string,
-): Promise<SchemaListEntry[]> {
+export async function cmdCasSchemaList(storageRoot: string): Promise<SchemaListEntry[]> {
   const store = openStore(storageRoot);
   const metaHash = await bootstrap(store);
   const entries: SchemaListEntry[] = [];
@@ -115,9 +106,7 @@ export async function cmdCasSchemaList(
   return entries;
 }
 
-export async function cmdCasReindex(
-  storageRoot: string,
-): Promise<{ status: string }> {
+export async function cmdCasReindex(storageRoot: string): Promise<{ status: string }> {
   const indexDir = join(storageRoot, "cas", "_index");
   const { rmSync } = await import("node:fs");
   rmSync(indexDir, { recursive: true, force: true });
@@ -126,10 +115,7 @@ export async function cmdCasReindex(
   return { status: "reindexed" };
 }
 
-export async function cmdCasSchemaGet(
-  storageRoot: string,
-  hash: string,
-): Promise<unknown> {
+export async function cmdCasSchemaGet(storageRoot: string, hash: string): Promise<unknown> {
   const store = openStore(storageRoot);
   const schema = getSchema(store, hash);
   if (schema === null) {
@@ -137,3 +123,10 @@ export async function cmdCasSchemaGet(
   }
   return schema;
 }
+
+export async function cmdCasPutText(storageRoot: string, text: string): Promise<{ hash: string }> {
+  const store = openStore(storageRoot);
+  const typeHash = await putSchema(store, TEXT_SCHEMA);
+  const hash = await store.put(typeHash, text);
+  return { hash };
+}

packages/cli-workflow/src/commands/log.ts

@@ -0,0 +1,116 @@
+import { readdir, readFile, stat, unlink } from "node:fs/promises";
+import { join } from "node:path";
+
+type LogListItem = {
+  name: string;
+  size: number;
+  date: string;
+};
+
+type LogShowFilter = {
+  thread: string | null;
+  process: string | null;
+  date: string | null;
+};
+
+type LogEntry = {
+  ts: string;
+  pid: string;
+  tag: string;
+  msg: string;
+  thread: string | null;
+  workflow: string | null;
+};
+
+type LogCleanResult = {
+  deleted: number;
+};
+
+function logsDir(storageRoot: string): string {
+  return join(storageRoot, "logs");
+}
+
+async function listLogFiles(dir: string): Promise<Array<string>> {
+  try {
+    const files = await readdir(dir);
+    return files.filter((f) => f.endsWith(".jsonl")).sort();
+  } catch {
+    return [];
+  }
+}
+
+function dateFromFilename(name: string): string {
+  return name.replace(".jsonl", "");
+}
+
+async function parseJsonlFile(path: string): Promise<Array<LogEntry>> {
+  const content = await readFile(path, "utf-8");
+  const lines = content
+    .trim()
+    .split("\n")
+    .filter((l) => l.length > 0);
+  return lines.map((line) => JSON.parse(line) as LogEntry);
+}
+
+export async function cmdLogList(storageRoot: string): Promise<Array<LogListItem>> {
+  const dir = logsDir(storageRoot);
+  const files = await listLogFiles(dir);
+  const items: Array<LogListItem> = [];
+  for (const name of files) {
+    const s = await stat(join(dir, name));
+    items.push({ name, size: s.size, date: dateFromFilename(name) });
+  }
+  // sort by date descending
+  items.sort((a, b) => (a.date > b.date ? -1 : a.date < b.date ? 1 : 0));
+  return items;
+}
+
+export async function cmdLogShow(
+  storageRoot: string,
+  filter: LogShowFilter,
+): Promise<Array<LogEntry>> {
+  const dir = logsDir(storageRoot);
+  let files: Array<string>;
+
+  if (filter.date !== null) {
+    files = [`${filter.date}.jsonl`];
+  } else {
+    files = await listLogFiles(dir);
+  }
+
+  let entries: Array<LogEntry> = [];
+  for (const file of files) {
+    try {
+      const parsed = await parseJsonlFile(join(dir, file));
+      entries = entries.concat(parsed);
+    } catch {
+      // file doesn't exist or is unreadable, skip
+    }
+  }
+
+  if (filter.thread !== null) {
+    entries = entries.filter((e) => e.thread === filter.thread);
+  }
+  if (filter.process !== null) {
+    entries = entries.filter((e) => e.pid === filter.process);
+  }
+
+  entries.sort((a, b) => (a.ts < b.ts ? -1 : a.ts > b.ts ? 1 : 0));
+  return entries;
+}
+
+export async function cmdLogClean(storageRoot: string, before: string): Promise<LogCleanResult> {
+  const dir = logsDir(storageRoot);
+  const files = await listLogFiles(dir);
+  let deleted = 0;
+
+  for (const name of files) {
+    const date = dateFromFilename(name);
+    if (date < before) {
+      await unlink(join(dir, name));
+      deleted++;
+    }
+  }
+
+  return { deleted };
+}

packages/cli-workflow/src/commands/setup.ts

@@ -1,10 +1,45 @@
-import { existsSync, readFileSync, writeFileSync, mkdirSync } from "node:fs";
-import { homedir } from "node:os";
-import { join, resolve } from "node:path";
-import { createInterface } from "node:readline/promises";
+import { existsSync, mkdirSync, readdirSync, readFileSync, statSync, writeFileSync } from "node:fs";
+import { join } from "node:path";
 import { stdin as input, stdout as output } from "node:process";
+import { createInterface } from "node:readline/promises";
+import type { Result } from "@uncaged/workflow-util";
+import { parse, stringify } from "yaml";
 
-import { stringify, parse } from "yaml";
+/**
+ * Send a minimal chat completion request to verify the model is reachable.
+ * Returns ok on 2xx, error with reason string otherwise.
+ */
+export async function validateModel(
+  baseUrl: string,
+  apiKey: string,
+  model: string,
+): Promise<Result<void, string>> {
+  try {
+    const url = `${baseUrl.replace(/\/+$/, "")}/chat/completions`;
+    const res = await fetch(url, {
+      method: "POST",
+      headers: {
+        Authorization: `Bearer ${apiKey}`,
+        "Content-Type": "application/json",
+      },
+      body: JSON.stringify({
+        model,
+        messages: [{ role: "user", content: "hi" }],
+        max_tokens: 1,
+      }),
+      signal: AbortSignal.timeout(15_000),
+    });
+    if (!res.ok) {
+      return { ok: false, error: `HTTP ${res.status} ${res.statusText}` };
+    }
+    return { ok: true, value: undefined };
+  } catch (err: unknown) {
+    if (err instanceof DOMException && err.name === "AbortError") {
+      return { ok: false, error: "Request timed out — model endpoint unreachable" };
+    }
+    return { ok: false, error: `Network error — could not reach endpoint (${String(err)})` };
+  }
+}
 
 /**
  * Preset provider list — embedded to avoid runtime YAML loading dependency.
@@ -17,10 +52,18 @@ const PRESET_PROVIDERS = [
   { name: "openrouter", label: "OpenRouter", baseUrl: "https://openrouter.ai/api/v1" },
   { name: "venice", label: "Venice", baseUrl: "https://api.venice.ai/api/v1" },
   // China
-  { name: "dashscope", label: "DashScope (Alibaba)", baseUrl: "https://dashscope.aliyuncs.com/compatible-mode/v1" },
+  {
+    name: "dashscope",
+    label: "DashScope (Alibaba)",
+    baseUrl: "https://dashscope.aliyuncs.com/compatible-mode/v1",
+  },
   { name: "deepseek", label: "DeepSeek", baseUrl: "https://api.deepseek.com/v1" },
   { name: "siliconflow", label: "SiliconFlow", baseUrl: "https://api.siliconflow.cn/v1" },
-  { name: "volcengine", label: "Volcengine (ByteDance)", baseUrl: "https://ark.cn-beijing.volces.com/api/v3" },
+  {
+    name: "volcengine",
+    label: "Volcengine (ByteDance)",
+    baseUrl: "https://ark.cn-beijing.volces.com/api/v3",
+  },
   { name: "kimi", label: "Kimi (Moonshot)", baseUrl: "https://api.moonshot.cn/v1" },
   { name: "glm", label: "GLM (Zhipu AI)", baseUrl: "https://open.bigmodel.cn/api/paas/v4" },
   { name: "stepfun", label: "StepFun", baseUrl: "https://api.stepfun.com/v1" },
@@ -94,25 +137,207 @@ function apiKeyEnvName(providerName: string): string {
   return `${providerName.toUpperCase().replace(/[^A-Z0-9]/g, "_")}_API_KEY`;
 }
 
+// ──────────────────────────────────────────────────────────────────────────────
+// Extracted helpers — _discoverAgents
+// ──────────────────────────────────────────────────────────────────────────────
+
+/**
+ * Scans directories from a PATH string for uwf-* executables.
+ */
+export async function _searchPathDirs(pathEnv: string): Promise<string[]> {
+  if (!pathEnv) return [];
+  const dirs = pathEnv.split(":").filter((d) => d.length > 0);
+  const agents = new Set<string>();
+  for (const dir of dirs) {
+    _scanDirForAgents(dir, agents);
+  }
+  return Array.from(agents).sort();
+}
+
+function _scanDirForAgents(dir: string, agents: Set<string>): void {
+  try {
+    if (!existsSync(dir)) return;
+    const entries = readdirSync(dir);
+    for (const entry of entries) {
+      if (!entry.startsWith("uwf-") || entry === "uwf") continue;
+      if (_isExecutableFile(join(dir, entry))) {
+        agents.add(entry);
+      }
+    }
+  } catch {
+    // Skip inaccessible directories
+  }
+}
+
+function _isExecutableFile(fullPath: string): boolean {
+  try {
+    const s = statSync(fullPath);
+    return s.isFile() && (s.mode & 0o111) !== 0;
+  } catch {
+    return false;
+  }
+}
+
+/**
+ * Parses the stdout of `which -a` into sorted unique basenames.
+ */
+export function _parseWhichOutput(text: string): string[] {
+  if (!text) return [];
+  const agents = new Set<string>();
+  for (const line of text.trim().split("\n")) {
+    if (!line) continue;
+    const basename = line.split("/").pop() ?? "";
+    if (basename.startsWith("uwf-") && basename !== "uwf") {
+      agents.add(basename);
+    }
+  }
+  return Array.from(agents).sort();
+}
+
+/**
+ * Discover uwf-* agent binaries in PATH.
+ * Returns sorted list of binary names (e.g., ["uwf-hermes", "uwf-claude-code"]).
+ */
+export async function _discoverAgents(): Promise<string[]> {
+  try {
+    const agents = await _tryWhichDiscovery();
+    if (agents !== null) return agents;
+    return await _searchPathDirs(process.env.PATH ?? "");
+  } catch {
+    return [];
+  }
+}
+
+async function _tryWhichDiscovery(): Promise<string[] | null> {
+  try {
+    const proc = Bun.spawn(["which", "-a", "uwf-hermes", "uwf-claude-code", "uwf-cursor"], {
+      stdout: "pipe",
+      stderr: "pipe",
+    });
+    const text = await new Response(proc.stdout).text();
+    await proc.exited;
+    if (proc.exitCode !== 0) return null;
+    return _parseWhichOutput(text);
+  } catch {
+    return null;
+  }
+}
+
+// ──────────────────────────────────────────────────────────────────────────────
+// Extracted helpers — onData closure (promptSecret)
+// ──────────────────────────────────────────────────────────────────────────────
+
+/** Returns true for newline, carriage return, or EOF (EOT). */
+export function _isTerminator(c: string): boolean {
+  return c === "\n" || c === "\r" || c === "";
+}
+
+/** Returns true for DEL or backspace. */
+export function _isBackspace(c: string): boolean {
+  return c === "" || c === "\b";
+}
+
+// ──────────────────────────────────────────────────────────────────────────────
+// Extracted helpers — cmdSetupInteractive
+// ──────────────────────────────────────────────────────────────────────────────
+
+type ProviderEntry = { name: string; label: string; baseUrl: string };
+
+/** Prints the numbered provider list and custom option to stdout. */
+export function _printProviderMenu(providers: readonly ProviderEntry[]): void {
+  const numWidth = String(providers.length + 1).length;
+  for (let i = 0; i < providers.length; i++) {
+    const p = providers[i];
+    if (!p) continue;
+    const num = String(i + 1).padStart(numWidth);
+    console.log(`  ${num}) ${p.label.padEnd(28)} ${p.baseUrl}`);
+  }
+  const customNum = String(providers.length + 1).padStart(numWidth);
+  console.log(`  ${customNum}) Custom (enter name and URL manually)\n`);
+}
+
+/** Resolves a numeric choice string to a preset provider, or null for custom/invalid. */
+export function _resolveProviderChoice(
+  choice: string,
+  providers: readonly ProviderEntry[],
+): { providerName: string; baseUrl: string } | null {
+  const n = Number.parseInt(choice, 10);
+  if (Number.isNaN(n) || n < 1 || n > providers.length) return null;
+  const p = providers[n - 1];
+  if (!p) return null;
+  return { providerName: p.name, baseUrl: p.baseUrl };
+}
+
+/** Resolves numeric index or literal model name to a model string. */
+export function _resolveModelChoice(input: string, models: string[]): string {
+  const n = Number.parseInt(input, 10);
+  if (!Number.isNaN(n) && n >= 1 && n <= models.length) {
+    return models[n - 1] ?? input;
+  }
+  return input;
+}
+
+/** Prints the multi-column model list to stdout. */
+export function _printModelMenu(models: string[], termCols: number): void {
+  const nw = String(models.length).length;
+  const maxLen = models.reduce((m, s) => Math.max(m, s.length), 0);
+  const colWidth = nw + 2 + maxLen + 4;
+  const cols = Math.max(1, Math.floor(termCols / colWidth));
+  const rows = Math.ceil(models.length / cols);
+  for (let r = 0; r < rows; r++) {
+    let line = "";
+    for (let c = 0; c < cols; c++) {
+      const idx = c * rows + r;
+      if (idx >= models.length) break;
+      const num = String(idx + 1).padStart(nw);
+      const name = (models[idx] ?? "").padEnd(maxLen);
+      line += `  ${num}) ${name}  `;
+    }
+    console.log(line.trimEnd());
+  }
+}
+
+type ValidationResult = { ok: boolean; error: string | null };
+
+/** Prints the model validation result to stdout. */
+export function _printValidationResult(validation: ValidationResult): void {
+  if (validation.ok) {
+    console.log("✓ Model verified — connection successful.\n");
+  } else {
+    console.log(`\n⚠ Warning: Could not reach model — ${validation.error}`);
+    console.log(
+      "  Config saved, but you may want to try a different model or check your API key.\n",
+    );
+  }
+}
+
+// ──────────────────────────────────────────────────────────────────────────────
+
 /**
  * Merge setup args into config.yaml structure. Non-destructive — preserves existing entries.
  */
 function mergeConfig(existing: Record<string, unknown>, args: SetupArgs): Record<string, unknown> {
-  const providers = (typeof existing.providers === "object" && existing.providers !== null
-    ? { ...(existing.providers as Record<string, unknown>) }
-    : {}) as Record<string, unknown>;
+  const providers = (
+    typeof existing.providers === "object" && existing.providers !== null
+      ? { ...(existing.providers as Record<string, unknown>) }
+      : {}
+  ) as Record<string, unknown>;
 
   const envName = apiKeyEnvName(args.provider);
   providers[args.provider] = { baseUrl: args.baseUrl, apiKeyEnv: envName };
 
-  const models = (typeof existing.models === "object" && existing.models !== null
-    ? { ...(existing.models as Record<string, unknown>) }
-    : {}) as Record<string, unknown>;
+  const models = (
+    typeof existing.models === "object" && existing.models !== null
+      ? { ...(existing.models as Record<string, unknown>) }
+      : {}
+  ) as Record<string, unknown>;
   models.default = { provider: args.provider, name: args.model };
 
-  const agents = (typeof existing.agents === "object" && existing.agents !== null
-    ? { ...(existing.agents as Record<string, unknown>) }
-    : {}) as Record<string, unknown>;
+  const agents = (
+    typeof existing.agents === "object" && existing.agents !== null
+      ? { ...(existing.agents as Record<string, unknown>) }
+      : {}
+  ) as Record<string, unknown>;
 
   const agentName = args.agent ?? "hermes";
   if (Object.keys(agents).length === 0) {
@@ -150,15 +375,59 @@ export async function cmdSetup(args: SetupArgs): Promise<Record<string, unknown>
   envData[envName] = args.apiKey;
   saveEnvFile(envPath, envData);
 
+  // Validate model connectivity
+  const validation = await validateModel(args.baseUrl, args.apiKey, args.model);
+
   return {
     configPath,
     envPath,
     provider: args.provider,
     model: args.model,
     defaultAgent: merged.defaultAgent,
+    validation,
   };
 }
 
+type SecretState = {
+  buf: string;
+  rawWasSet: boolean;
+  resolve: (value: string) => void;
+  onData: (chunk: string) => void;
+};
+
+function _handleSecretTerminator(state: SecretState): void {
+  if (process.stdin.isTTY) process.stdin.setRawMode(state.rawWasSet);
+  process.stdin.pause();
+  process.stdin.removeListener("data", state.onData);
+  process.stdout.write("\n");
+  state.resolve(state.buf.trim());
+}
+
+function _handleSecretBackspace(state: SecretState): void {
+  if (state.buf.length > 0) {
+    state.buf = state.buf.slice(0, -1);
+    process.stdout.write("\b \b");
+  }
+}
+
+function _handleSecretChar(c: string, state: SecretState): boolean {
+  if (_isTerminator(c)) {
+    _handleSecretTerminator(state);
+    return true;
+  }
+  if (_isBackspace(c)) {
+    _handleSecretBackspace(state);
+    return false;
+  }
+  if (c === "") {
+    if (process.stdin.isTTY) process.stdin.setRawMode(state.rawWasSet);
+    process.exit(130);
+  }
+  state.buf += c;
+  process.stdout.write("*");
+  return false;
+}
+
 /** Read a line with terminal echo disabled (for secrets). */
 async function promptSecret(label: string): Promise<string> {
   process.stdout.write(label);
@@ -170,33 +439,13 @@ async function promptSecret(label: string): Promise<string> {
     process.stdin.resume();
     process.stdin.setEncoding("utf8");
 
-    let buf = "";
-    const onData = (chunk: string) => {
+    const state: SecretState = { buf: "", rawWasSet, resolve, onData: () => {} };
+    state.onData = (chunk: string) => {
       for (const c of chunk.toString()) {
-        if (c === "\n" || c === "\r" || c === "\u0004") {
-          if (process.stdin.isTTY) process.stdin.setRawMode(rawWasSet);
-          process.stdin.pause();
-          process.stdin.removeListener("data", onData);
-          process.stdout.write("\n");
-          resolve(buf.trim());
-          return;
-        }
-        if (c === "\u007F" || c === "\b") {
-          if (buf.length > 0) {
-            buf = buf.slice(0, -1);
-            process.stdout.write("\b \b");
-          }
-          continue;
-        }
-        if (c === "\u0003") {
-          if (process.stdin.isTTY) process.stdin.setRawMode(rawWasSet);
-          process.exit(130);
-        }
-        buf += c;
-        process.stdout.write("*");
+        if (_handleSecretChar(c, state)) return;
       }
     };
-    process.stdin.on("data", onData);
+    process.stdin.on("data", state.onData);
   });
 }
 
@@ -211,13 +460,67 @@ async function fetchModels(baseUrl: string, apiKey: string): Promise<string[]> {
     if (!res.ok) return [];
     const body = (await res.json()) as { data?: { id: string }[] };
     if (!Array.isArray(body.data)) return [];
-    const NON_CHAT = /speech|embed|image|video|audio|ocr|rerank|tts|asr|paraformer|sambert|cosyvoice|wordart|wanx|wan2|flux|stable-diffusion|gui-/i;
-    return body.data.map((m) => m.id).filter((id) => !NON_CHAT.test(id)).sort();
+    const NON_CHAT =
+      /speech|embed|image|video|audio|ocr|rerank|tts|asr|paraformer|sambert|cosyvoice|wordart|wanx|wan2|flux|stable-diffusion|gui-/i;
+    return body.data
+      .map((m) => m.id)
+      .filter((id) => !NON_CHAT.test(id))
+      .sort();
   } catch {
     return [];
   }
 }
 
+async function _promptProviderSelection(
+  rl: ReturnType<typeof createInterface>,
+): Promise<{ providerName: string; baseUrl: string }> {
+  console.log("Select a provider:\n");
+  _printProviderMenu(PRESET_PROVIDERS);
+
+  const choice = (await rl.question(`Choose [1-${PRESET_PROVIDERS.length + 1}]: `)).trim();
+  const choiceNum = Number.parseInt(choice, 10);
+  if (Number.isNaN(choiceNum) || choiceNum < 1 || choiceNum > PRESET_PROVIDERS.length + 1) {
+    throw new Error(`Invalid choice: ${choice}`);
+  }
+
+  const preset = _resolveProviderChoice(choice, PRESET_PROVIDERS);
+  if (preset) {
+    const selected = PRESET_PROVIDERS[choiceNum - 1];
+    if (selected) {
+      console.log(`\n  → ${selected.label} (${selected.baseUrl})\n`);
+    }
+    return preset;
+  }
+
+  const providerName = (await rl.question("Provider name (e.g. my-proxy): ")).trim();
+  if (!providerName) throw new Error("Provider name required");
+  const baseUrl = (await rl.question("OpenAI-compatible API base URL: ")).trim();
+  if (!baseUrl) throw new Error("Base URL required");
+  return { providerName, baseUrl };
+}
+
+async function _promptModelSelection(
+  rl: ReturnType<typeof createInterface>,
+  baseUrl: string,
+  apiKey: string,
+): Promise<string> {
+  console.log("\nFetching available models...");
+  const models = await fetchModels(baseUrl, apiKey);
+
+  if (models.length === 0) {
+    console.log("Could not fetch models. Enter model name manually.");
+    const model = (await rl.question("Default model (e.g. qwen-plus, gpt-4o): ")).trim();
+    if (!model) throw new Error("Model required");
+    return model;
+  }
+  console.log(`\nAvailable models (${models.length}):\n`);
+  _printModelMenu(models, process.stdout.columns || 100);
+  console.log(`\nChoose a number, or type a model name directly.`);
+  const modelInput = (await rl.question(`Default model [1-${models.length}]: `)).trim();
+  if (!modelInput) throw new Error("Model required");
+  return _resolveModelChoice(modelInput, models);
+}
+
 /**
  * Interactive setup — prompts user for provider, API key, model.
  */
@@ -227,39 +530,7 @@ export async function cmdSetupInteractive(storageRoot: string): Promise<Record<s
   try {
     console.log("Configure LLM provider for uwf workflow agents.\n");
 
-    // 1. Provider selection
-    const numWidth = String(PRESET_PROVIDERS.length + 1).length;
-    console.log("Select a provider:\n");
-    for (let i = 0; i < PRESET_PROVIDERS.length; i++) {
-      const p = PRESET_PROVIDERS[i];
-      if (!p) continue;
-      const num = String(i + 1).padStart(numWidth);
-      console.log(`  ${num}) ${p.label.padEnd(28)} ${p.baseUrl}`);
-    }
-    const customNum = String(PRESET_PROVIDERS.length + 1).padStart(numWidth);
-    console.log(`  ${customNum}) Custom (enter name and URL manually)\n`);
-
-    const choice = (await rl.question(`Choose [1-${PRESET_PROVIDERS.length + 1}]: `)).trim();
-    const choiceNum = Number.parseInt(choice, 10);
-    if (Number.isNaN(choiceNum) || choiceNum < 1 || choiceNum > PRESET_PROVIDERS.length + 1) {
-      throw new Error(`Invalid choice: ${choice}`);
-    }
-
-    let providerName: string;
-    let baseUrl: string;
-
-    if (choiceNum <= PRESET_PROVIDERS.length) {
-      const selected = PRESET_PROVIDERS[choiceNum - 1];
-      if (!selected) throw new Error("Invalid selection");
-      providerName = selected.name;
-      baseUrl = selected.baseUrl;
-      console.log(`\n  → ${selected.label} (${selected.baseUrl})\n`);
-    } else {
-      providerName = (await rl.question("Provider name (e.g. my-proxy): ")).trim();
-      if (!providerName) throw new Error("Provider name required");
-      baseUrl = (await rl.question("OpenAI-compatible API base URL: ")).trim();
-      if (!baseUrl) throw new Error("Base URL required");
-    }
+    const { providerName, baseUrl } = await _promptProviderSelection(rl);
 
     // 2. API key
     rl.close();
@@ -268,50 +539,11 @@ export async function cmdSetupInteractive(storageRoot: string): Promise<Record<s
 
     // 3. Model selection
     const rl2 = createInterface({ input, output });
-    console.log("\nFetching available models...");
-    const models = await fetchModels(baseUrl, apiKey);
-
-    let model: string;
-    if (models.length > 0) {
-      console.log(`\nAvailable models (${models.length}):\n`);
-      const nw = String(models.length).length;
-      // Multi-column layout
-      const maxLen = models.reduce((m, s) => Math.max(m, s.length), 0);
-      const colWidth = nw + 2 + maxLen + 4; // "  N) name    "
-      const termCols = process.stdout.columns || 100;
-      const cols = Math.max(1, Math.floor(termCols / colWidth));
-      const rows = Math.ceil(models.length / cols);
-      for (let r = 0; r < rows; r++) {
-        let line = "";
-        for (let c = 0; c < cols; c++) {
-          const idx = c * rows + r;
-          if (idx >= models.length) break;
-          const num = String(idx + 1).padStart(nw);
-          const name = (models[idx] ?? "").padEnd(maxLen);
-          line += `  ${num}) ${name}  `;
-        }
-        console.log(line.trimEnd());
-      }
-      console.log(`\nChoose a number, or type a model name directly.`);
-      const modelInput = (await rl2.question(`Default model [1-${models.length}]: `)).trim();
-      if (!modelInput) throw new Error("Model required");
-      const modelNum = Number.parseInt(modelInput, 10);
-      if (!Number.isNaN(modelNum) && modelNum >= 1 && modelNum <= models.length) {
-        model = models[modelNum - 1] ?? modelInput;
-      } else {
-        model = modelInput;
-      }
-    } else {
-      console.log("Could not fetch models. Enter model name manually.");
-      model = (await rl2.question("Default model (e.g. qwen-plus, gpt-4o): ")).trim();
-      if (!model) throw new Error("Model required");
-    }
-
+    const model = await _promptModelSelection(rl2, baseUrl, apiKey);
     rl2.close();
-
     console.log(`  → ${providerName}/${model}\n`);
 
-    await cmdSetup({
+    const setupResult = await cmdSetup({
       provider: providerName,
       baseUrl,
       apiKey,
@@ -319,6 +551,10 @@ export async function cmdSetupInteractive(storageRoot: string): Promise<Record<s
       storageRoot,
     });
 
+    // Show validation result
+    if (setupResult.validation && typeof setupResult.validation === "object") {
+      _printValidationResult(setupResult.validation as ValidationResult);
+    }
     console.log("Setup complete! Get started:\n");
     console.log("  uwf workflow put <workflow.yaml>   Register a workflow");
     console.log('  uwf thread start <name> -p "..."   Start a thread');

packages/cli-workflow/src/commands/skill.ts

@@ -0,0 +1 @@
+export { generateCliReference as cmdSkillCli } from "@uncaged/workflow-util";

packages/cli-workflow/src/commands/thread.ts

@@ -1,4 +1,6 @@
-import { execFileSync } from "node:child_process";
+import { execFileSync, spawn } from "node:child_process";
+import { access, readFile } from "node:fs/promises";
+import { dirname, isAbsolute, resolve as resolvePath } from "node:path";
 import type { Store as CasStore, JSONSchema } from "@uncaged/json-cas";
 import { getSchema, validate } from "@uncaged/json-cas";
 import { getEnvPath, loadWorkflowConfig } from "@uncaged/workflow-agent-kit";
@@ -8,6 +10,7 @@ import type {
   AgentConfig,
   CasRef,
   ModeratorContext,
+  RunningThreadsOutput,
   StartEntry,
   StartNodePayload,
   StartOutput,
@@ -22,10 +25,15 @@ import type {
   WorkflowConfig,
   WorkflowPayload,
 } from "@uncaged/workflow-protocol";
-import { generateUlid } from "@uncaged/workflow-util";
+import { createProcessLogger, generateUlid, type ProcessLogger } from "@uncaged/workflow-util";
 import { config as loadDotenv } from "dotenv";
-import { stringify } from "yaml";
-
+import { parse, stringify } from "yaml";
+import {
+  createMarker,
+  deleteMarker,
+  isThreadRunning,
+  listRunningThreads,
+} from "../background/index.js";
 import {
   appendThreadHistory,
   createUwfStore,
@@ -38,11 +46,25 @@ import {
   type ThreadHistoryLine,
   type UwfStore,
 } from "../store.js";
-import { isCasRef } from "../validate.js";
+import { checkWorkflowFilenameConsistency, isCasRef, parseWorkflowPayload } from "../validate.js";
+import { materializeWorkflowPayload } from "./workflow.js";
 
 const END_ROLE = "$END";
 export const THREAD_READ_DEFAULT_QUOTA = 4000;
 
+const PL_THREAD_START = "7HNQ4B2X";
+const PL_MODERATOR = "M3K8V9T1";
+const PL_AGENT_SPAWN = "R5J2W8N4";
+const PL_AGENT_DONE = "C6P9E3H7";
+const PL_THREAD_ARCHIVED = "F4D8Q2K5";
+const PL_STEP_ERROR = "B8T5N1V6";
+const PL_BACKGROUND_START = "X7Q4W9M2";
+
+function failStep(plog: ProcessLogger, message: string): never {
+  plog.log(PL_STEP_ERROR, message, null);
+  fail(message);
+}
+
 type ChainState = {
   startHash: CasRef;
   start: StartNodePayload;
@@ -66,15 +88,159 @@ function fail(message: string): never {
   process.exit(1);
 }
 
+/**
+ * Check if a string looks like a file path (contains path separators or has .yaml/.yml extension).
+ */
+function isFilePath(input: string): boolean {
+  return (
+    input.includes("/") || input.includes("\\") || input.endsWith(".yaml") || input.endsWith(".yml")
+  );
+}
+
+/**
+ * Check if a workflow file exists at the given path.
+ */
+async function workflowFileExists(dir: string, name: string, ext: string): Promise<string | null> {
+  const candidate = resolvePath(dir, `${name}${ext}`);
+  try {
+    await access(candidate);
+    return candidate;
+  } catch {
+    return null;
+  }
+}
+
+/**
+ * Search for a workflow file in a given directory (checks both .workflow/ and .workflows/).
+ */
+async function findWorkflowInDir(dir: string, name: string): Promise<string | null> {
+  // Check .workflow/ directory first (preferred)
+  for (const ext of [".yaml", ".yml"]) {
+    const result = await workflowFileExists(resolvePath(dir, ".workflow"), name, ext);
+    if (result !== null) {
+      return result;
+    }
+  }
+
+  // Check .workflows/ directory as fallback (legacy)
+  for (const ext of [".yaml", ".yml"]) {
+    const result = await workflowFileExists(resolvePath(dir, ".workflows"), name, ext);
+    if (result !== null) {
+      return result;
+    }
+  }
+
+  return null;
+}
+
+/**
+ * Traverse parent directories looking for `.workflow/<name>.yaml` or `.workflow/<name>.yml`.
+ * Returns the absolute path if found, otherwise null.
+ * Stops at filesystem root or .git directory.
+ */
+async function findWorkflowInParents(startDir: string, name: string): Promise<string | null> {
+  let currentDir = resolvePath(startDir);
+  const root = resolvePath("/");
+
+  while (true) {
+    const found = await findWorkflowInDir(currentDir, name);
+    if (found !== null) {
+      return found;
+    }
+
+    // Stop at filesystem root
+    if (currentDir === root) {
+      break;
+    }
+
+    // Move to parent directory
+    const parentDir = dirname(currentDir);
+    if (parentDir === currentDir) {
+      // Reached filesystem root
+      break;
+    }
+    currentDir = parentDir;
+  }
+
+  return null;
+}
+
+async function materializeLocalWorkflow(uwf: UwfStore, filePath: string): Promise<CasRef> {
+  let text: string;
+  try {
+    text = await readFile(filePath, "utf8");
+  } catch {
+    fail(`project workflow file not found: ${filePath}`);
+  }
+
+  let raw: unknown;
+  try {
+    raw = parse(text) as unknown;
+  } catch (e) {
+    fail(`invalid YAML in ${filePath}: ${e instanceof Error ? e.message : String(e)}`);
+  }
+
+  const payload = parseWorkflowPayload(raw);
+  if (payload === null) {
+    fail(`invalid workflow YAML in ${filePath}: expected WorkflowPayload shape`);
+  }
+
+  const filenameError = checkWorkflowFilenameConsistency(filePath, payload);
+  if (filenameError !== null) {
+    fail(filenameError);
+  }
+
+  const materialized = await materializeWorkflowPayload(uwf, payload);
+  const hash = await uwf.store.put(uwf.schemas.workflow, materialized);
+  const stored = uwf.store.get(hash);
+  if (stored === null || !validate(uwf.store, stored)) {
+    fail("stored local workflow failed schema validation");
+  }
+
+  return hash;
+}
+
 async function resolveWorkflowCasRef(
   uwf: UwfStore,
   storageRoot: string,
   workflowId: string,
+  projectRoot: string,
 ): Promise<CasRef> {
+  // Validate input
+  const trimmed = workflowId.trim();
+  if (trimmed === "") {
+    fail("workflow ID cannot be empty");
+  }
+
+  // Strategy 1: Direct CAS hash
+  if (isCasRef(trimmed)) {
+    const node = uwf.store.get(trimmed);
+    if (node === null) {
+      fail(`CAS node not found: ${trimmed}`);
+    }
+    if (node.type !== uwf.schemas.workflow) {
+      fail(`node ${trimmed} is not a Workflow (type ${node.type})`);
+    }
+    return trimmed;
+  }
+
+  // Strategy 2: Explicit file path (relative or absolute)
+  if (isFilePath(trimmed)) {
+    const absolutePath = isAbsolute(trimmed) ? trimmed : resolvePath(projectRoot, trimmed);
+    return materializeLocalWorkflow(uwf, absolutePath);
+  }
+
+  // Strategy 3: Local discovery (parent directory traversal)
+  const localPath = await findWorkflowInParents(projectRoot, trimmed);
+  if (localPath !== null) {
+    return materializeLocalWorkflow(uwf, localPath);
+  }
+
+  // Strategy 4: Global registry fallback
   const registry = await loadWorkflowRegistry(storageRoot);
-  const hash = resolveWorkflowHash(registry, workflowId);
+  const hash = resolveWorkflowHash(registry, trimmed);
   if (!isCasRef(hash)) {
-    fail(`workflow not found: ${workflowId}`);
+    fail(`workflow not found: ${trimmed}`);
   }
   const node = uwf.store.get(hash);
   if (node === null) {
@@ -114,11 +280,16 @@ export async function cmdThreadStart(
   storageRoot: string,
   workflowId: string,
   prompt: string,
+  projectRoot: string,
 ): Promise<StartOutput> {
   const uwf = await createUwfStore(storageRoot);
-  const workflowHash = await resolveWorkflowCasRef(uwf, storageRoot, workflowId);
+  const workflowHash = await resolveWorkflowCasRef(uwf, storageRoot, workflowId, projectRoot);
 
   const threadId = generateUlid(Date.now()) as ThreadId;
+  const plog = createProcessLogger({
+    storageRoot,
+    context: { thread: threadId, workflow: workflowHash },
+  });
   const startPayload: StartNodePayload = {
     workflow: workflowHash,
     prompt,
@@ -134,6 +305,12 @@ export async function cmdThreadStart(
   index[threadId] = headHash;
   await saveThreadsIndex(storageRoot, index);
 
+  plog.log(
+    PL_THREAD_START,
+    `thread created workflow=${workflowHash} thread=${threadId} head=${headHash}`,
+    null,
+  );
+
   return { workflow: workflowHash, thread: threadId };
 }
 
@@ -151,6 +328,7 @@ export async function cmdThreadShow(storageRoot: string, threadId: ThreadId): Pr
       thread: threadId,
       head: activeHead,
       done: false,
+      background: null,
     };
   }
 
@@ -161,6 +339,7 @@ export async function cmdThreadShow(storageRoot: string, threadId: ThreadId): Pr
       thread: threadId,
       head: hist.head,
       done: true,
+      background: null,
     };
   }
 
@@ -292,49 +471,68 @@ function expandDeep(store: CasStore, hash: CasRef, visited?: Set<string>): unkno
   return expandValue(store, schema, node.payload, seen);
 }
 
+function expandCasRefField(store: CasStore, value: unknown, visited: Set<string>): unknown {
+  if (typeof value === "string") {
+    return expandDeep(store, value as CasRef, visited);
+  }
+  return value;
+}
+
+function expandAnyOfField(
+  store: CasStore,
+  schema: JSONSchema,
+  value: unknown,
+  visited: Set<string>,
+): unknown {
+  if (!Array.isArray(schema.anyOf)) return value;
+  for (const sub of schema.anyOf as JSONSchema[]) {
+    if (sub.format === "cas_ref" && typeof value === "string") {
+      return expandDeep(store, value as CasRef, visited);
+    }
+  }
+  return value;
+}
+
+function expandArrayField(
+  store: CasStore,
+  schema: JSONSchema,
+  value: unknown,
+  visited: Set<string>,
+): unknown {
+  if (!schema.items || !Array.isArray(value)) return value;
+  const itemSchema = schema.items as JSONSchema;
+  return (value as unknown[]).map((item) => expandValue(store, itemSchema, item, visited));
+}
+
+function expandObjectField(
+  store: CasStore,
+  schema: JSONSchema,
+  value: unknown,
+  visited: Set<string>,
+): unknown {
+  if (value === null || typeof value !== "object" || Array.isArray(value) || !schema.properties) {
+    return value;
+  }
+  const props = schema.properties as Record<string, JSONSchema>;
+  const obj = value as Record<string, unknown>;
+  const result: Record<string, unknown> = {};
+  for (const [key, val] of Object.entries(obj)) {
+    const propSchema = props[key];
+    result[key] = propSchema ? expandValue(store, propSchema, val, visited) : val;
+  }
+  return result;
+}
+
 function expandValue(
   store: CasStore,
   schema: JSONSchema,
   value: unknown,
   visited: Set<string>,
 ): unknown {
-  // If this field is a cas_ref, expand it
-  if (schema.format === "cas_ref") {
-    if (typeof value === "string") {
-      return expandDeep(store, value as CasRef, visited);
-    }
-    return value;
-  }
-
-  // anyOf (nullable refs)
-  if (Array.isArray(schema.anyOf)) {
-    for (const sub of schema.anyOf as JSONSchema[]) {
-      if (sub.format === "cas_ref" && typeof value === "string") {
-        return expandDeep(store, value as CasRef, visited);
-      }
-    }
-    return value;
-  }
-
-  // Array of cas_ref items
-  if (schema.type === "array" && schema.items && Array.isArray(value)) {
-    const itemSchema = schema.items as JSONSchema;
-    return (value as unknown[]).map((item) => expandValue(store, itemSchema, item, visited));
-  }
-
-  // Object with properties
-  if (value !== null && typeof value === "object" && !Array.isArray(value) && schema.properties) {
-    const props = schema.properties as Record<string, JSONSchema>;
-    const obj = value as Record<string, unknown>;
-    const result: Record<string, unknown> = {};
-    for (const [key, val] of Object.entries(obj)) {
-      const propSchema = props[key];
-      result[key] = propSchema ? expandValue(store, propSchema, val, visited) : val;
-    }
-    return result;
-  }
-
-  return value;
+  if (schema.format === "cas_ref") return expandCasRefField(store, value, visited);
+  if (Array.isArray(schema.anyOf)) return expandAnyOfField(store, schema, value, visited);
+  if (schema.type === "array") return expandArrayField(store, schema, value, visited);
+  return expandObjectField(store, schema, value, visited);
 }
 
 function collectOrderedSteps(
@@ -369,7 +567,7 @@ function collectOrderedSteps(
 }
 
 function formatYaml(value: unknown): string {
-  return stringify(value).trimEnd();
+  return stringify(value, { aliasDuplicateObjects: false }).trimEnd();
 }
 
 function formatCompactStep(index: number, item: OrderedStepItem, outputYaml: string): string {
@@ -418,6 +616,85 @@ export function extractLastAssistantContent(uwf: UwfStore, detailRef: CasRef): s
   return null;
 }
 
+function sliceBeforeHash(
+  candidates: OrderedStepItem[],
+  before: CasRef,
+  threadId: ThreadId,
+): OrderedStepItem[] {
+  const idx = candidates.findIndex((s) => s.hash === before);
+  if (idx === -1) {
+    fail(`step ${before} not found in thread ${threadId}`);
+  }
+  return candidates.slice(0, idx);
+}
+
+function selectByQuota(
+  candidates: OrderedStepItem[],
+  uwf: UwfStore,
+  quota: number,
+): { selected: OrderedStepItem[]; skippedCount: number } {
+  const selected: OrderedStepItem[] = [];
+  let totalChars = 0;
+  for (let i = candidates.length - 1; i >= 0; i--) {
+    const item = candidates[i];
+    if (item === undefined) continue;
+    const outputYaml = formatYaml(expandOutput(uwf, item.payload.output));
+    const blockLen = formatCompactStep(i + 1, item, outputYaml).length;
+    selected.unshift(item);
+    totalChars += blockLen;
+    if (totalChars > quota) break;
+  }
+  return { selected, skippedCount: candidates.length - selected.length };
+}
+
+function formatStepHeader(stepNum: number, item: OrderedStepItem): string {
+  const ts = new Date(item.timestamp)
+    .toISOString()
+    .replace("T", " ")
+    .replace(/\.\d+Z$/, "");
+  return [
+    `## Step ${stepNum}: ${item.payload.role} \`${item.hash}\``,
+    `**Agent:** ${item.payload.agent} | **Time:** ${ts}`,
+  ].join("\n");
+}
+
+function formatStepPrompt(
+  roleDef: WorkflowPayload["roles"][string] | undefined,
+  role: string,
+  shownPromptRoles: Set<string>,
+): string {
+  if (!roleDef || shownPromptRoles.has(role)) return "";
+  shownPromptRoles.add(role);
+  return ["", "", "<prompt>", roleDef.goal, "</prompt>"].join("\n");
+}
+
+function formatStepContent(uwf: UwfStore, item: OrderedStepItem): string {
+  if (!item.payload.detail) return "";
+  const content = extractLastAssistantContent(uwf, item.payload.detail);
+  if (content === null) return "";
+  return ["", "", "<output>", content, "</output>"].join("\n");
+}
+
+function formatStartSection(options: {
+  threadId: ThreadId;
+  workflowName: string;
+  workflowHash: CasRef;
+  prompt: string;
+  before: CasRef | null;
+  showStart: boolean;
+}): string {
+  if (options.before !== null && !options.showStart) return "";
+  return [
+    `# Thread \`${options.threadId}\``,
+    "",
+    `**Workflow:** ${options.workflowName} (\`${options.workflowHash}\`)`,
+    "",
+    "## Task",
+    "",
+    options.prompt,
+  ].join("\n");
+}
+
 function formatThreadReadMarkdown(options: {
   threadId: ThreadId;
   workflowName: string;
@@ -430,50 +707,16 @@ function formatThreadReadMarkdown(options: {
   before: CasRef | null;
   showStart: boolean;
 }): string {
-  const { ordered, uwf, workflow, quota, before, showStart } = options;
+  const { ordered, uwf, workflow, quota, before } = options;
 
-  // Determine which steps to consider
-  let candidates = ordered;
-  if (before !== null) {
-    const idx = candidates.findIndex((s) => s.hash === before);
-    if (idx === -1) {
-      fail(`step ${before} not found in thread ${options.threadId}`);
-    }
-    candidates = candidates.slice(0, idx);
-  }
+  const candidates = before !== null ? sliceBeforeHash(ordered, before, options.threadId) : ordered;
+  const { selected, skippedCount } = selectByQuota(candidates, uwf, quota);
 
-  // Walk backward from newest, accumulating chars until quota exceeded
-  const selected: OrderedStepItem[] = [];
-  let totalChars = 0;
-  for (let i = candidates.length - 1; i >= 0; i--) {
-    const item = candidates[i];
-    if (item === undefined) continue;
-    const outputYaml = formatYaml(expandOutput(uwf, item.payload.output));
-    const blockLen = formatCompactStep(i + 1, item, outputYaml).length;
-    selected.unshift(item);
-    totalChars += blockLen;
-    if (totalChars > quota) break;
-  }
-
-  const skippedCount = candidates.length - selected.length;
   const parts: string[] = [];
 
-  // Start section
-  if (before === null || showStart) {
-    parts.push(
-      [
-        `# Thread \`${options.threadId}\``,
-        "",
-        `**Workflow:** ${options.workflowName} (\`${options.workflowHash}\`)`,
-        "",
-        "## Task",
-        "",
-        options.prompt,
-      ].join("\n"),
-    );
-  }
+  const startSection = formatStartSection(options);
+  if (startSection !== "") parts.push(startSection);
 
-  // Skip hint
   if (skippedCount > 0 && selected.length > 0) {
     const firstSelected = selected[0];
     if (firstSelected !== undefined) {
@@ -483,33 +726,21 @@ function formatThreadReadMarkdown(options: {
     }
   }
 
-  // Step blocks
   const startIndex = candidates.length - selected.length;
+  const shownPromptRoles = new Set<string>();
   for (let i = 0; i < selected.length; i++) {
     const item = selected[i];
     if (item === undefined) continue;
     const stepNum = startIndex + i + 1;
-    const outputYaml = formatYaml(expandOutput(uwf, item.payload.output));
-    const ts = new Date(item.timestamp)
-      .toISOString()
-      .replace("T", " ")
-      .replace(/\.\d+Z$/, "");
-    const stepLines = [
-      `## Step ${stepNum}: ${item.payload.role} \`${item.hash}\``,
-      `**Agent:** ${item.payload.agent} | **Time:** ${ts}`,
-    ];
     const roleDef = workflow.roles[item.payload.role];
-    if (roleDef) {
-      stepLines.push("", "### Prompt", "", roleDef.systemPrompt);
-    }
-    if (item.payload.detail) {
-      const content = extractLastAssistantContent(uwf, item.payload.detail);
-      if (content !== null) {
-        stepLines.push("", "### Content", "", content);
-      }
-    }
-    stepLines.push("", "### Output", "", "```yaml", outputYaml, "```");
-    parts.push(stepLines.join("\n"));
+    const stepBlock = [
+      formatStepHeader(stepNum, item),
+      formatStepPrompt(roleDef, item.payload.role, shownPromptRoles),
+      formatStepContent(uwf, item),
+    ]
+      .filter((s) => s !== "")
+      .join("");
+    parts.push(stepBlock);
   }
 
   return parts.join("\n\n---\n\n");
@@ -522,6 +753,7 @@ function buildModeratorContext(uwf: UwfStore, chain: ChainState): ModeratorConte
     output: expandOutput(uwf, step.output),
     detail: step.detail,
     agent: step.agent,
+    edgePrompt: step.edgePrompt ?? "",
   }));
   return { start: chain.start, steps };
 }
@@ -574,30 +806,36 @@ function resolveAgentConfig(
   return agentConfig;
 }
 
-function spawnAgent(agent: AgentConfig, threadId: ThreadId, role: string): CasRef {
-  const argv = [...agent.args, threadId, role];
+function spawnAgent(
+  plog: ProcessLogger,
+  agent: AgentConfig,
+  threadId: ThreadId,
+  role: string,
+  edgePrompt: string,
+): CasRef {
+  const argv = [...agent.args, "--thread", threadId, "--role", role, "--prompt", edgePrompt];
   let stdout: string;
   try {
     stdout = execFileSync(agent.command, argv, {
       encoding: "utf8",
-      env: process.env,
       stdio: ["ignore", "pipe", "pipe"],
+      maxBuffer: 50 * 1024 * 1024, // 50 MB — stream-json output can be large
     });
   } catch (e) {
-    const err = e as NodeJS.ErrnoException & { stderr?: Buffer | string };
+    const err = e as NodeJS.ErrnoException & { stderr?: Buffer | string | null };
     const stderr =
-      err.stderr === undefined
+      err.stderr == null
         ? ""
         : typeof err.stderr === "string"
           ? err.stderr
           : err.stderr.toString("utf8");
     const detail = stderr.trim() !== "" ? `: ${stderr.trim()}` : "";
-    fail(`agent command failed (${agent.command})${detail}`);
+    failStep(plog, `agent command failed (${agent.command})${detail}`);
   }
 
   const line = stdout.trim().split("\n").pop()?.trim() ?? "";
   if (!isCasRef(line)) {
-    fail(`agent stdout is not a valid CAS hash: ${line || "(empty)"}`);
+    failStep(plog, `agent stdout is not a valid CAS hash: ${line || "(empty)"}`);
   }
   return line;
 }
@@ -623,12 +861,139 @@ export async function cmdThreadStep(
   storageRoot: string,
   threadId: ThreadId,
   agentOverride: string | null,
-): Promise<StepOutput> {
+  count: number,
+  background: boolean,
+  backgroundWorker: boolean,
+): Promise<StepOutput[]> {
+  if (count < 1 || !Number.isInteger(count)) {
+    fail(`--count must be a positive integer, got: ${count}`);
+  }
+
+  // Check if thread is already running in background (unless we ARE the background worker)
+  if (!backgroundWorker) {
+    const runningMarker = await isThreadRunning(storageRoot, threadId);
+    if (runningMarker !== null) {
+      fail(`thread already executing in background (PID: ${runningMarker.pid})`);
+    }
+  }
+
+  const workflowHash = await resolveActiveThreadWorkflowHash(storageRoot, threadId);
+  const plog = createProcessLogger({
+    storageRoot,
+    context: { thread: threadId, workflow: workflowHash },
+  });
+
+  if (background && !backgroundWorker) {
+    // Spawn background process
+    return cmdThreadStepBackground(storageRoot, threadId, agentOverride, count, plog, workflowHash);
+  }
+
+  // If we're the background worker, create marker before execution
+  let markerCreated = false;
+  if (backgroundWorker) {
+    await createMarker(storageRoot, {
+      thread: threadId,
+      workflow: workflowHash,
+      pid: process.pid,
+      startedAt: Date.now(),
+    });
+    markerCreated = true;
+  }
+
+  try {
+    const results: StepOutput[] = [];
+    for (let i = 0; i < count; i++) {
+      const result = await cmdThreadStepOnce(storageRoot, threadId, agentOverride, plog);
+      results.push(result);
+      if (result.done) {
+        break;
+      }
+    }
+    return results;
+  } finally {
+    // Cleanup marker if we created one
+    if (markerCreated) {
+      await deleteMarker(storageRoot, threadId);
+    }
+  }
+}
+
+async function resolveActiveThreadWorkflowHash(
+  storageRoot: string,
+  threadId: ThreadId,
+): Promise<CasRef> {
   const index = await loadThreadsIndex(storageRoot);
   const headHash = index[threadId];
   if (headHash === undefined) {
     fail(`thread not active: ${threadId}`);
   }
+  const uwf = await createUwfStore(storageRoot);
+  const chain = walkChain(uwf, headHash);
+  return chain.start.workflow;
+}
+
+async function cmdThreadStepBackground(
+  storageRoot: string,
+  threadId: ThreadId,
+  agentOverride: string | null,
+  count: number,
+  plog: ProcessLogger,
+  workflowHash: CasRef,
+): Promise<StepOutput[]> {
+  // Get current head to return to caller
+  const index = await loadThreadsIndex(storageRoot);
+  const headHash = index[threadId];
+  if (headHash === undefined) {
+    failStep(plog, `thread not active: ${threadId}`);
+  }
+
+  // Spawn detached background process
+  const scriptPath = process.argv[1];
+  if (scriptPath === undefined) {
+    failStep(plog, "unable to determine script path for background execution");
+  }
+
+  const args = ["thread", "step", threadId, "--count", String(count)];
+
+  if (agentOverride !== null) {
+    args.push("--agent", agentOverride);
+  }
+
+  // Internal flag to signal the background worker to create/cleanup markers
+  args.push("--_background-worker");
+
+  plog.log(PL_BACKGROUND_START, `spawning background process count=${count}`, null);
+
+  const child = spawn(scriptPath, args, {
+    detached: true,
+    stdio: "ignore",
+  });
+
+  child.unref();
+
+  // Return immediately with current state and background flag
+  return [
+    {
+      workflow: workflowHash,
+      thread: threadId,
+      head: headHash,
+      done: false,
+      background: true,
+    },
+  ];
+}
+
+async function cmdThreadStepOnce(
+  storageRoot: string,
+  threadId: ThreadId,
+  agentOverride: string | null,
+  plog: ProcessLogger,
+): Promise<StepOutput> {
+  const index = await loadThreadsIndex(storageRoot);
+  const headHash = index[threadId];
+  if (headHash === undefined) {
+    failStep(plog, `thread not active: ${threadId}`);
+  }
 
   const uwf = await createUwfStore(storageRoot);
   const chain = walkChain(uwf, headHash);
@@ -638,31 +1003,46 @@ export async function cmdThreadStep(
 
   const nextResult = await evaluate(workflow, context);
   if (!nextResult.ok) {
-    fail(nextResult.error.message);
+    failStep(plog, `moderator evaluate failed: ${nextResult.error.message}`);
   }
 
-  if (nextResult.value === END_ROLE) {
+  plog.log(
+    PL_MODERATOR,
+    `moderator role=${nextResult.value.role} prompt=${nextResult.value.prompt}`,
+    null,
+  );
+
+  if (nextResult.value.role === END_ROLE) {
+    plog.log(PL_THREAD_ARCHIVED, `thread archived head=${headHash}`, null);
     await archiveThread(storageRoot, threadId, workflowHash, headHash);
     return {
       workflow: workflowHash,
       thread: threadId,
       head: headHash,
       done: true,
+      background: null,
     };
   }
 
-  const role = nextResult.value;
+  const role = nextResult.value.role;
+  const edgePrompt = nextResult.value.prompt;
   const config = await loadWorkflowConfig(storageRoot);
   const agent = resolveAgentConfig(config, workflow, role, agentOverride);
 
+  plog.log(PL_AGENT_SPAWN, `spawning agent command=${agent.command}`, {
+    args: [...agent.args, threadId, role].join(" "),
+  });
+
   loadDotenv({ path: getEnvPath(storageRoot) });
-  const newHead = spawnAgent(agent, threadId, role);
+  const newHead = spawnAgent(plog, agent, threadId, role, edgePrompt);
+
+  plog.log(PL_AGENT_DONE, `agent returned head=${newHead}`, null);
 
   // Re-create store to pick up nodes written by the agent subprocess
   const uwfAfter = await createUwfStore(storageRoot);
   const newNode = uwfAfter.store.get(newHead);
   if (newNode === null || newNode.type !== uwfAfter.schemas.stepNode) {
-    fail(`agent returned hash that is not a StepNode: ${newHead}`);
+    failStep(plog, `agent returned hash that is not a StepNode: ${newHead}`);
   }
 
   // Reload threads index to avoid overwriting changes made by the agent subprocess
@@ -674,11 +1054,12 @@ export async function cmdThreadStep(
   const contextAfter = buildModeratorContext(uwfAfter, chainAfter);
   const afterResult = await evaluate(workflow, contextAfter);
   if (!afterResult.ok) {
-    fail(afterResult.error.message);
+    failStep(plog, `post-step moderator evaluate failed: ${afterResult.error.message}`);
   }
 
-  const done = afterResult.value === END_ROLE;
+  const done = afterResult.value.role === END_ROLE;
   if (done) {
+    plog.log(PL_THREAD_ARCHIVED, `thread archived head=${newHead}`, null);
     await archiveThread(storageRoot, threadId, workflowHash, newHead);
   }
 
@@ -687,6 +1068,7 @@ export async function cmdThreadStep(
     thread: threadId,
     head: newHead,
     done,
+    background: null,
   };
 }
 
@@ -823,6 +1205,17 @@ export async function cmdThreadKill(storageRoot: string, threadId: ThreadId): Pr
     fail(`thread not active: ${threadId}`);
   }
 
+  // Check if thread is running in background and terminate it
+  const runningMarker = await isThreadRunning(storageRoot, threadId);
+  if (runningMarker !== null) {
+    try {
+      process.kill(runningMarker.pid, "SIGTERM");
+    } catch {
+      // Process may have already exited, ignore error
+    }
+    await deleteMarker(storageRoot, threadId);
+  }
+
   const uwf = await createUwfStore(storageRoot);
   const workflow = resolveWorkflowFromHead(uwf, head);
   if (workflow === null) {
@@ -842,3 +1235,8 @@ export async function cmdThreadKill(storageRoot: string, threadId: ThreadId): Pr
 
   return { thread: threadId, archived: true };
 }
+
+export async function cmdThreadRunning(storageRoot: string): Promise<RunningThreadsOutput> {
+  const threads = await listRunningThreads(storageRoot);
+  return { threads };
+}

packages/cli-workflow/src/commands/workflow.ts

@@ -2,22 +2,31 @@ import { readFile } from "node:fs/promises";
 
 import type { JSONSchema } from "@uncaged/json-cas";
 import { putSchema, validate } from "@uncaged/json-cas";
-import type { CasRef, RoleDefinition, WorkflowPayload } from "@uncaged/workflow-protocol";
+import type {
+  CasRef,
+  RoleDefinition,
+  Transition,
+  WorkflowPayload,
+} from "@uncaged/workflow-protocol";
 import { parse } from "yaml";
 
 import {
   createUwfStore,
+  discoverProjectWorkflows,
   findRegistryName,
   loadWorkflowRegistry,
   resolveWorkflowHash,
   saveWorkflowRegistry,
   type UwfStore,
 } from "../store.js";
-import { parseWorkflowPayload } from "../validate.js";
+import { checkWorkflowFilenameConsistency, parseWorkflowPayload } from "../validate.js";
+
+export type WorkflowOrigin = "local" | "global";
 
 export type WorkflowListEntry = {
   name: string;
   hash: CasRef;
+  origin: WorkflowOrigin;
 };
 
 export type WorkflowPutOutput = {
@@ -42,35 +51,55 @@ function isJsonSchema(value: unknown): value is JSONSchema {
   return typeof value === "object" && value !== null && !Array.isArray(value);
 }
 
-async function resolveOutputSchemaRef(
+/** Normalize graph transitions: ensure condition is null (not undefined) for fallback entries. */
+function normalizeGraph(graph: Record<string, Transition[]>): Record<string, Transition[]> {
+  const result: Record<string, Transition[]> = {};
+  for (const [node, transitions] of Object.entries(graph)) {
+    result[node] = transitions.map((t) => {
+      if (typeof t.prompt !== "string" || t.prompt.trim() === "") {
+        fail(`graph[${node}] transition to "${t.role}": prompt is required (non-empty string)`);
+      }
+      return {
+        role: t.role,
+        condition: t.condition ?? null,
+        prompt: t.prompt,
+      };
+    });
+  }
+  return result;
+}
+
+async function resolveFrontmatterRef(
   uwf: UwfStore,
   roleName: string,
-  outputSchema: unknown,
+  frontmatter: unknown,
 ): Promise<CasRef> {
-  if (!isJsonSchema(outputSchema)) {
-    fail(`role "${roleName}": outputSchema must be a JSON Schema object`);
+  if (!isJsonSchema(frontmatter)) {
+    fail(`role "${roleName}": frontmatter must be a JSON Schema object`);
   }
-  const schema: JSONSchema = outputSchema.title === undefined
-    ? { ...outputSchema, title: roleName }
-    : outputSchema;
+  const schema: JSONSchema =
+    frontmatter.title === undefined ? { ...frontmatter, title: roleName } : frontmatter;
   return putSchema(uwf.store, schema);
 }
 
-async function materializeWorkflowPayload(
+export async function materializeWorkflowPayload(
   uwf: UwfStore,
   raw: WorkflowPayload,
 ): Promise<WorkflowPayload> {
   const roles: Record<string, RoleDefinition> = {};
   for (const [roleName, role] of Object.entries(raw.roles)) {
-    const outputSchema = await resolveOutputSchemaRef(
+    const frontmatter = await resolveFrontmatterRef(
       uwf,
       `${raw.name}.${roleName}`,
-      role.outputSchema,
+      role.frontmatter,
     );
     roles[roleName] = {
       description: role.description,
-      systemPrompt: role.systemPrompt,
-      outputSchema,
+      goal: role.goal,
+      capabilities: role.capabilities,
+      procedure: role.procedure,
+      output: role.output,
+      frontmatter,
     };
   }
   return {
@@ -78,7 +107,7 @@ async function materializeWorkflowPayload(
     description: raw.description,
     roles,
     conditions: raw.conditions,
-    graph: raw.graph,
+    graph: normalizeGraph(raw.graph),
   };
 }
 
@@ -105,6 +134,11 @@ export async function cmdWorkflowPut(
     fail("invalid workflow YAML: expected WorkflowPayload shape");
   }
 
+  const filenameError = checkWorkflowFilenameConsistency(filePath, payload);
+  if (filenameError !== null) {
+    fail(filenameError);
+  }
+
   const uwf = await createUwfStore(storageRoot);
   const materialized = await materializeWorkflowPayload(uwf, payload);
 
@@ -147,7 +181,26 @@ export async function cmdWorkflowShow(
   };
 }
 
-export async function cmdWorkflowList(storageRoot: string): Promise<WorkflowListEntry[]> {
+export async function cmdWorkflowList(
+  storageRoot: string,
+  projectRoot: string,
+): Promise<WorkflowListEntry[]> {
+  const localEntries = await discoverProjectWorkflows(projectRoot);
   const registry = await loadWorkflowRegistry(storageRoot);
-  return Object.entries(registry).map(([name, hash]) => ({ name, hash }));
+
+  const result: WorkflowListEntry[] = [];
+  const localNames = new Set<string>();
+
+  for (const entry of localEntries) {
+    localNames.add(entry.name);
+    result.push({ name: entry.name, hash: "(local)", origin: "local" });
+  }
+
+  for (const [name, hash] of Object.entries(registry)) {
+    if (!localNames.has(name)) {
+      result.push({ name, hash, origin: "global" });
+    }
+  }
+
+  return result;
 }

packages/cli-workflow/src/format.ts

@@ -7,6 +7,6 @@ export function formatOutput(data: unknown, format: OutputFormat): string {
     case "json":
       return JSON.stringify(data);
     case "yaml":
-      return stringify(data).trimEnd();
+      return stringify(data, { aliasDuplicateObjects: false }).trimEnd();
   }
 }

packages/cli-workflow/src/schemas.ts

@@ -1,15 +1,14 @@
 import type { Hash, Store } from "@uncaged/json-cas";
 import { putSchema } from "@uncaged/json-cas";
-import {
-  START_NODE_SCHEMA,
-  STEP_NODE_SCHEMA,
-  WORKFLOW_SCHEMA,
-} from "@uncaged/workflow-protocol";
+import { START_NODE_SCHEMA, STEP_NODE_SCHEMA, WORKFLOW_SCHEMA } from "@uncaged/workflow-protocol";
+
+export const TEXT_SCHEMA = { type: "string" as const };
 
 export type UwfSchemaHashes = {
   workflow: Hash;
   startNode: Hash;
   stepNode: Hash;
+  text: Hash;
 };
 
 /**
@@ -17,10 +16,11 @@ export type UwfSchemaHashes = {
  * Idempotent: safe to call on every CLI invocation.
  */
 export async function registerUwfSchemas(store: Store): Promise<UwfSchemaHashes> {
-  const [workflow, startNode, stepNode] = await Promise.all([
+  const [workflow, startNode, stepNode, text] = await Promise.all([
     putSchema(store, WORKFLOW_SCHEMA),
     putSchema(store, START_NODE_SCHEMA),
     putSchema(store, STEP_NODE_SCHEMA),
+    putSchema(store, TEXT_SCHEMA),
   ]);
-  return { workflow, startNode, stepNode };
+  return { workflow, startNode, stepNode, text };
 }

packages/cli-workflow/src/store.ts

@@ -1,8 +1,8 @@
-import { appendFile, mkdir, readFile, writeFile } from "node:fs/promises";
+import { appendFile, mkdir, readdir, readFile, writeFile } from "node:fs/promises";
 import { homedir } from "node:os";
 import { join } from "node:path";
 
-import type { Hash, Store } from "@uncaged/json-cas";
+import type { BootstrapCapableStore, Hash } from "@uncaged/json-cas";
 import { createFsStore } from "@uncaged/json-cas-fs";
 import type { CasRef, ThreadId, ThreadListItem, ThreadsIndex } from "@uncaged/workflow-protocol";
 import { parse, stringify } from "yaml";
@@ -11,6 +11,44 @@ import { registerUwfSchemas, type UwfSchemaHashes } from "./schemas.js";
 
 export type WorkflowRegistry = Record<string, CasRef>;
 
+/** A workflow entry discovered from the project-local .workflows/ directory. */
+export type ProjectWorkflowEntry = {
+  /** Workflow name (from YAML `name` field, equals filename stem). */
+  name: string;
+  /** Absolute path to the YAML file. */
+  filePath: string;
+};
+
+/**
+ * Scan `<projectRoot>/.workflows/*.yaml` (non-recursive) and return discovered entries.
+ * Returns an empty array if the directory does not exist.
+ */
+export async function discoverProjectWorkflows(
+  projectRoot: string,
+): Promise<ProjectWorkflowEntry[]> {
+  const dir = join(projectRoot, ".workflows");
+  let entries: string[];
+  try {
+    entries = await readdir(dir);
+  } catch (e) {
+    const err = e as NodeJS.ErrnoException;
+    if (err.code === "ENOENT" || err.code === "ENOTDIR") {
+      return [];
+    }
+    throw e;
+  }
+
+  const result: ProjectWorkflowEntry[] = [];
+  for (const entry of entries) {
+    if (!entry.endsWith(".yaml") && !entry.endsWith(".yml")) {
+      continue;
+    }
+    const stem = entry.endsWith(".yaml") ? entry.slice(0, -5) : entry.slice(0, -4);
+    result.push({ name: stem, filePath: join(dir, entry) });
+  }
+  return result;
+}
+
 /** Default filesystem root for uwf data (`~/.uncaged/workflow`). */
 export function getDefaultStorageRoot(): string {
   return join(homedir(), ".uncaged", "workflow");
@@ -54,7 +92,7 @@ export type ThreadHistoryLine = ThreadListItem & {
 
 export type UwfStore = {
   storageRoot: string;
-  store: Store;
+  store: BootstrapCapableStore;
   schemas: UwfSchemaHashes;
 };
 
@@ -104,6 +142,22 @@ export function resolveWorkflowHash(registry: WorkflowRegistry, id: string): Cas
   return registry[id] !== undefined ? registry[id] : id;
 }
 
+/**
+ * Resolve a workflow name to a project-local YAML file path.
+ * Returns null if the name is not found in the local entries.
+ */
+export function resolveProjectWorkflowFile(
+  localEntries: ProjectWorkflowEntry[],
+  name: string,
+): string | null {
+  for (const entry of localEntries) {
+    if (entry.name === name) {
+      return entry.filePath;
+    }
+  }
+  return null;
+}
+
 export function findRegistryName(registry: WorkflowRegistry, hash: Hash): string | null {
   for (const [name, h] of Object.entries(registry)) {
     if (h === hash) {

packages/cli-workflow/src/validate.ts

@@ -1,3 +1,4 @@
+import { basename } from "node:path";
 import type { CasRef, WorkflowPayload } from "@uncaged/workflow-protocol";
 
 const CAS_REF_PATTERN = /^[0-9A-HJKMNP-TV-Z]{13}$/;
@@ -14,10 +15,18 @@ function isRoleDefinition(value: unknown): boolean {
   if (!isRecord(value)) {
     return false;
   }
-  const outputSchema = value.outputSchema;
-  const schemaOk = isRecord(outputSchema) && typeof outputSchema.type === "string";
+  const frontmatter = value.frontmatter;
+  const frontmatterOk = isRecord(frontmatter) && typeof frontmatter.type === "string";
+  const capabilities = value.capabilities;
+  const capabilitiesOk =
+    Array.isArray(capabilities) && capabilities.every((c) => typeof c === "string");
   return (
-    typeof value.description === "string" && typeof value.systemPrompt === "string" && schemaOk
+    typeof value.description === "string" &&
+    typeof value.goal === "string" &&
+    capabilitiesOk &&
+    typeof value.procedure === "string" &&
+    typeof value.output === "string" &&
+    frontmatterOk
   );
 }
 
@@ -33,7 +42,12 @@ function isTransition(value: unknown): boolean {
     return false;
   }
   const condition = value.condition;
-  return typeof value.role === "string" && (condition === null || typeof condition === "string");
+  return (
+    typeof value.role === "string" &&
+    typeof value.prompt === "string" &&
+    value.prompt.trim() !== "" &&
+    (condition === null || condition === undefined || typeof condition === "string")
+  );
 }
 
 function isStringRecord(value: unknown, itemCheck: (item: unknown) => boolean): boolean {
@@ -52,6 +66,33 @@ function isGraph(value: unknown): boolean {
   );
 }
 
+/**
+ * Derive the expected workflow name from a file path (stem without extension).
+ * Returns the stem for `.yaml` / `.yml` files.
+ */
+export function workflowNameFromPath(filePath: string): string {
+  const base = basename(filePath);
+  if (base.endsWith(".yaml")) return base.slice(0, -5);
+  if (base.endsWith(".yml")) return base.slice(0, -4);
+  return base;
+}
+
+/**
+ * Check that the `name` field in a parsed payload matches the expected name
+ * derived from the file path.  Returns an error message string on mismatch,
+ * or null when the names are consistent.
+ */
+export function checkWorkflowFilenameConsistency(
+  filePath: string,
+  payload: WorkflowPayload,
+): string | null {
+  const expected = workflowNameFromPath(filePath);
+  if (payload.name !== expected) {
+    return `workflow name mismatch: file "${basename(filePath)}" implies name "${expected}" but YAML declares name "${payload.name}"`;
+  }
+  return null;
+}
+
 /** Validate YAML-parsed workflow document shape (outputSchema may be inline JSON Schema). */
 export function parseWorkflowPayload(raw: unknown): WorkflowPayload | null {
   if (!isRecord(raw)) {

packages/workflow-agent-builtin/README.md

@@ -0,0 +1,141 @@
+# @uncaged/workflow-agent-builtin
+
+`uwf-builtin` agent — built-in LLM agent with file read/write and shell tools.
+
+## Overview
+
+Layer 3 agent implementation. Runs an OpenAI-compatible chat completion loop with built-in tools (`read_file`, `write_file`, `run_command`). Uses the configured provider/model from `config.yaml`. Produces frontmatter markdown output and stores turn-by-turn session detail in CAS.
+
+Useful when you want a self-contained agent without an external CLI like Hermes or Claude Code.
+
+**Dependencies:** `@uncaged/json-cas`, `@uncaged/workflow-agent-kit`, `@uncaged/workflow-util`
+
+## Installation
+
+Included as the `uwf-builtin` binary when you install `@uncaged/workflow-agent-builtin`:
+
+```bash
+bun add -g @uncaged/workflow-agent-builtin
+```
+
+## CLI Usage
+
+Invoked by `uwf thread step`:
+
+```bash
+uwf-builtin <thread-id> <role>
+```
+
+Configure as default agent:
+
+```bash
+uwf setup --agent builtin
+```
+
+Override per step:
+
+```bash
+uwf thread step <thread-id> --agent uwf-builtin
+```
+
+Environment variables set by the engine:
+
+| Variable | Purpose |
+|----------|---------|
+| `UWF_EDGE_PROMPT` | Moderator edge instruction for this step |
+
+## API
+
+All exports come from `src/index.ts`.
+
+### Agent factory
+
+```typescript
+function createBuiltinAgent(): () => Promise<void>
+function buildBuiltinMessages(ctx: AgentContext): ChatMessage[]
+```
+
+### LLM loop
+
+```typescript
+const BUILTIN_MAX_TURNS = 30;
+const BUILTIN_CONTINUE_MAX_TURNS = 5;
+
+function runBuiltinLoop(/* options: RunBuiltinLoopOptions */): Promise<RunBuiltinLoopResult>
+function chatCompletionWithTools(
+  provider: ResolvedLlmProvider,
+  messages: ChatMessage[],
+  tools: OpenAiToolDefinition[],
+): Promise<LlmAssistantResponse>
+```
+
+`RunBuiltinLoopOptions` and `RunBuiltinLoopResult` are internal to `loop.ts` and not re-exported from `index.ts`.
+
+### Tools
+
+```typescript
+function getBuiltinTools(): readonly BuiltinTool[]
+function executeBuiltinTool(
+  name: string,
+  args: Record<string, unknown>,
+  ctx: ToolContext,
+): Promise<string>
+```
+
+### Session and detail
+
+```typescript
+function initSessionDir(storageRoot: string): Promise<void>
+function appendSessionTurn(storageRoot: string, sessionId: string, turn: BuiltinTurnPayload): Promise<void>
+function readSessionTurns(storageRoot: string, sessionId: string): Promise<BuiltinTurnPayload[]>
+function removeSession(storageRoot: string, sessionId: string): Promise<void>
+function registerBuiltinSchemas(store: Store): Promise<BuiltinSchemaHashes>
+function storeBuiltinDetail(store: Store, payload: BuiltinDetailPayload): Promise<string>
+```
+
+### Types
+
+```typescript
+type ChatMessage = /* system | user | assistant | tool */;
+type LlmAssistantResponse = { content: string | null; toolCalls: LlmToolCall[] | null };
+type LlmToolCall = { id: string; name: string; arguments: string };
+type BuiltinTool = { name: string; description: string; parameters: Record<string, unknown> };
+type ToolContext = { cwd: string; storageRoot: string };
+type BuiltinDetailPayload = { /* session turns, model, timestamps */ };
+type BuiltinLoopTurn = { /* single loop iteration record */ };
+type BuiltinToolCallRecord = { /* tool call audit */ };
+type BuiltinToolResultRecord = { /* tool result audit */ };
+type BuiltinTurnPayload = { /* persisted turn */ };
+```
+
+## Internal Structure
+
+```
+src/
+├── index.ts
+├── cli.ts              Binary entrypoint
+├── agent.ts            createBuiltinAgent
+├── loop.ts             Multi-turn LLM + tool loop
+├── prompt.ts           buildBuiltinMessages
+├── session.ts          Session directory persistence
+├── detail.ts           CAS detail node storage
+├── schemas.ts          Builtin CAS schemas
+├── types.ts            Detail and turn payload types
+├── llm/
+│   ├── index.ts
+│   ├── llm.ts          chatCompletionWithTools
+│   └── types.ts        ChatMessage, LlmToolCall, etc.
+└── tools/
+    ├── index.ts        getBuiltinTools, executeBuiltinTool
+    ├── read-file.ts
+    ├── write-file.ts
+    ├── run-command.ts
+    ├── path.ts
+    └── types.ts
+```
+
+## Configuration
+
+Requires a configured OpenAI-compatible provider and model in `~/.uncaged/workflow/config.yaml` (via `uwf setup`). API keys are loaded from `~/.uncaged/workflow/.env`.
+
+Tools run with the current working directory as `ToolContext.cwd` (typically the directory where `uwf thread step` was invoked).

packages/workflow-agent-builtin/__tests__/llm-parse.test.ts

@@ -0,0 +1,16 @@
+import { describe, expect, test } from "bun:test";
+
+import type { LlmToolCall } from "../src/llm/types.js";
+
+/** Mirror OpenAI response shape for parser coverage via chatCompletionWithTools integration later. */
+describe("LlmToolCall shape", () => {
+  test("tool call record fields", () => {
+    const call: LlmToolCall = {
+      id: "call_1",
+      name: "read_file",
+      arguments: '{"path":"README.md"}',
+    };
+    expect(call.name).toBe("read_file");
+    expect(JSON.parse(call.arguments)).toEqual({ path: "README.md" });
+  });
+});

packages/workflow-agent-builtin/__tests__/loop.test.ts

@@ -0,0 +1,256 @@
+import { beforeEach, describe, expect, mock, test } from "bun:test";
+
+const mockChatCompletionWithTools = mock(async () => ({
+  content: "---\nstatus: done\n---",
+  toolCalls: [],
+}));
+const mockAppendSessionTurn = mock(async () => {});
+const mockExecuteBuiltinTool = mock(async () => "tool-result");
+
+mock.module("../src/llm/index.js", () => ({
+  chatCompletionWithTools: mockChatCompletionWithTools,
+}));
+mock.module("../src/session.js", () => ({
+  appendSessionTurn: mockAppendSessionTurn,
+}));
+mock.module("../src/tools/index.js", () => ({
+  builtinToolsToOpenAi: () => [],
+  executeBuiltinTool: mockExecuteBuiltinTool,
+  getBuiltinTools: () => [],
+}));
+
+import {
+  executeTurnTools,
+  extractFinalText,
+  runBuiltinLoop,
+  shouldInjectDeadlineWarning,
+  shouldNudge,
+  shouldProcessToolCalls,
+} from "../src/loop.js";
+
+const fakeProvider = {} as any;
+const fakeToolCtx = {} as any;
+
+function makeOptions(overrides: Partial<Parameters<typeof runBuiltinLoop>[0]> = {}) {
+  return {
+    provider: fakeProvider,
+    messages: [{ role: "system" as const, content: "sys" }],
+    toolCtx: fakeToolCtx,
+    maxTurns: 5,
+    storageRoot: "/tmp",
+    sessionId: "sess",
+    noTools: false,
+    ...overrides,
+  };
+}
+
+beforeEach(() => {
+  mockChatCompletionWithTools.mockReset();
+  mockAppendSessionTurn.mockReset();
+  mockExecuteBuiltinTool.mockReset();
+});
+
+describe("shouldNudge", () => {
+  test("2.1 returns true when all conditions met", () => {
+    expect(shouldNudge({ noTools: false, text: "some text", turn: 0, maxTurns: 5 })).toBe(true);
+  });
+  test("2.2 returns false when noTools=true", () => {
+    expect(shouldNudge({ noTools: true, text: "some text", turn: 0, maxTurns: 5 })).toBe(false);
+  });
+  test("2.3 returns false when text starts with ---", () => {
+    expect(shouldNudge({ noTools: false, text: "---\nstatus: done", turn: 0, maxTurns: 5 })).toBe(
+      false,
+    );
+  });
+  test("2.4 returns false on last turn", () => {
+    expect(shouldNudge({ noTools: false, text: "some text", turn: 4, maxTurns: 5 })).toBe(false);
+  });
+  test("2.5 returns true on second-to-last turn", () => {
+    expect(shouldNudge({ noTools: false, text: "some text", turn: 3, maxTurns: 5 })).toBe(true);
+  });
+  test("2.6 leading whitespace before --- suppresses nudge", () => {
+    expect(shouldNudge({ noTools: false, text: "  ---\nstatus: done", turn: 0, maxTurns: 5 })).toBe(
+      false,
+    );
+  });
+});
+
+describe("executeTurnTools", () => {
+  test("4.1 executes each tool call and pushes tool result messages", async () => {
+    mockExecuteBuiltinTool.mockResolvedValue("result");
+    const messages: any[] = [];
+    const calls = [
+      { id: "c1", name: "tool_a", arguments: "{}" },
+      { id: "c2", name: "tool_b", arguments: "{}" },
+    ];
+    const count = await executeTurnTools(calls, fakeToolCtx, messages, "/tmp", "sess");
+    expect(messages.length).toBe(2);
+    expect(messages[0].role).toBe("tool");
+    expect(messages[1].role).toBe("tool");
+    expect(count).toBe(2);
+  });
+  test("4.2 tool result content matches executeBuiltinTool return value", async () => {
+    mockExecuteBuiltinTool.mockResolvedValue("result-A");
+    const messages: any[] = [];
+    await executeTurnTools(
+      [{ id: "c1", name: "read_file", arguments: "{}" }],
+      fakeToolCtx,
+      messages,
+      "/tmp",
+      "sess",
+    );
+    expect(messages[0].content).toBe("result-A");
+  });
+});
+
+describe("runBuiltinLoop integration", () => {
+  test("3.1 single text-only response returns finalText immediately", async () => {
+    mockChatCompletionWithTools.mockResolvedValue({
+      content: "---\nstatus: done\n---",
+      toolCalls: [],
+    });
+    const result = await runBuiltinLoop(makeOptions());
+    expect(result.finalText).toBe("---\nstatus: done\n---");
+    expect(result.turnCount).toBe(1);
+  });
+  test("3.2 noTools=true suppresses tool calls", async () => {
+    mockChatCompletionWithTools.mockResolvedValue({
+      content: "ok",
+      toolCalls: [{ id: "c1", name: "read_file", arguments: "{}" }],
+    });
+    const result = await runBuiltinLoop(makeOptions({ noTools: true }));
+    expect(result.finalText).toBe("ok");
+    expect(result.turnCount).toBe(1);
+  });
+  test("3.3 tool call followed by text response", async () => {
+    mockChatCompletionWithTools
+      .mockResolvedValueOnce({
+        content: null,
+        toolCalls: [{ id: "c1", name: "read_file", arguments: "{}" }],
+      })
+      .mockResolvedValueOnce({ content: "---\nstatus: done\n---", toolCalls: [] });
+    mockExecuteBuiltinTool.mockResolvedValue("file contents");
+    const result = await runBuiltinLoop(makeOptions());
+    expect(result.finalText).toBe("---\nstatus: done\n---");
+    expect(result.turnCount).toBe(3);
+  });
+  test("3.4 nudge cycle inserts nudge message", async () => {
+    mockChatCompletionWithTools
+      .mockResolvedValueOnce({ content: "I am thinking", toolCalls: [] })
+      .mockResolvedValueOnce({ content: "---\nstatus: done\n---", toolCalls: [] });
+    const result = await runBuiltinLoop(makeOptions());
+    expect(result.finalText).toBe("---\nstatus: done\n---");
+    const nudgeMsg = result.messages.find(
+      (m) =>
+        m.role === "user" && typeof m.content === "string" && m.content.includes("frontmatter"),
+    );
+    expect(nudgeMsg).toBeDefined();
+  });
+  test("3.5 maxTurns exhaustion falls back to last assistant content", async () => {
+    mockChatCompletionWithTools.mockResolvedValue({ content: "still thinking", toolCalls: [] });
+    const result = await runBuiltinLoop(makeOptions({ maxTurns: 3 }));
+    expect(result.finalText).toBe("still thinking");
+  });
+  test("3.6 original messages array is not mutated", async () => {
+    mockChatCompletionWithTools.mockResolvedValue({
+      content: "---\nstatus: done\n---",
+      toolCalls: [],
+    });
+    const original = [{ role: "system" as const, content: "sys" }];
+    await runBuiltinLoop(makeOptions({ messages: original }));
+    expect(original.length).toBe(1);
+  });
+});
+
+describe("shouldInjectDeadlineWarning", () => {
+  test("5.1 returns true when turn count reaches warning threshold and not yet warned", () => {
+    expect(shouldInjectDeadlineWarning(7, 10, false, false)).toBe(true);
+  });
+  test("5.2 returns false when already warned", () => {
+    expect(shouldInjectDeadlineWarning(7, 10, true, false)).toBe(false);
+  });
+  test("5.3 returns false when noTools is true", () => {
+    expect(shouldInjectDeadlineWarning(7, 10, false, true)).toBe(false);
+  });
+  test("5.4 returns false when turns remaining > DEADLINE_WARNING_TURNS", () => {
+    expect(shouldInjectDeadlineWarning(5, 10, false, false)).toBe(false);
+  });
+  test("5.5 returns true when exactly at warning threshold", () => {
+    expect(shouldInjectDeadlineWarning(7, 10, false, false)).toBe(true);
+  });
+  test("5.6 returns false when turns remaining is 0", () => {
+    expect(shouldInjectDeadlineWarning(10, 10, false, false)).toBe(false);
+  });
+});
+
+describe("shouldProcessToolCalls", () => {
+  test("6.1 returns true when toolCalls present and noTools=false", () => {
+    expect(shouldProcessToolCalls([{ id: "x", name: "read", arguments: "{}" }], false)).toBe(true);
+  });
+  test("6.2 returns false when toolCalls is null", () => {
+    expect(shouldProcessToolCalls(null, false)).toBe(false);
+  });
+  test("6.3 returns false when toolCalls is empty array", () => {
+    expect(shouldProcessToolCalls([], false)).toBe(false);
+  });
+  test("6.4 returns false when noTools=true", () => {
+    expect(shouldProcessToolCalls([{ id: "x", name: "read", arguments: "{}" }], true)).toBe(false);
+  });
+  test("6.5 returns true when multiple tool calls present", () => {
+    expect(
+      shouldProcessToolCalls(
+        [
+          { id: "x1", name: "read", arguments: "{}" },
+          { id: "x2", name: "write", arguments: "{}" },
+        ],
+        false,
+      ),
+    ).toBe(true);
+  });
+});
+
+describe("extractFinalText", () => {
+  test("7.1 returns last assistant message content", () => {
+    const messages = [
+      { role: "system" as const, content: "sys", tool_calls: null },
+      { role: "assistant" as const, content: "first", tool_calls: null },
+      { role: "assistant" as const, content: "last", tool_calls: null },
+    ];
+    expect(extractFinalText(messages)).toBe("last");
+  });
+  test("7.2 returns empty string when no assistant messages", () => {
+    expect(extractFinalText([{ role: "system" as const, content: "sys", tool_calls: null }])).toBe(
+      "",
+    );
+  });
+  test("7.3 skips assistant messages with null content", () => {
+    const messages = [
+      { role: "assistant" as const, content: "first", tool_calls: null },
+      {
+        role: "assistant" as const,
+        content: null,
+        tool_calls: [{ id: "x", name: "t", arguments: "{}" }],
+      },
+      { role: "assistant" as const, content: "second", tool_calls: null },
+    ];
+    expect(extractFinalText(messages)).toBe("second");
+  });
+  test("7.4 skips assistant messages with empty content", () => {
+    const messages = [
+      { role: "assistant" as const, content: "first", tool_calls: null },
+      { role: "assistant" as const, content: "", tool_calls: null },
+      { role: "user" as const, content: "nudge", tool_calls: null },
+    ];
+    expect(extractFinalText(messages)).toBe("first");
+  });
+  test("7.5 handles empty messages array", () => {
+    expect(extractFinalText([])).toBe("");
+  });
+  test("7.6 handles messages with only user and system roles", () => {
+    const messages = [
+      { role: "system" as const, content: "sys", tool_calls: null },
+      { role: "user" as const, content: "query", tool_calls: null },
+    ];
+    expect(extractFinalText(messages)).toBe("");
+  });
+});

packages/workflow-agent-builtin/__tests__/path.test.ts

@@ -0,0 +1,21 @@
+import { describe, expect, test } from "bun:test";
+import { resolve } from "node:path";
+import { resolvePath } from "../src/tools/path.js";
+
+describe("resolvePath", () => {
+  test("resolves relative paths against cwd", () => {
+    const root = "/workspace/project";
+    const resolved = resolvePath(root, "src/foo.ts");
+    expect(resolved).toBe(resolve(root, "src/foo.ts"));
+  });
+
+  test("resolves absolute paths as-is", () => {
+    const resolved = resolvePath("/workspace", "/etc/hosts");
+    expect(resolved).toBe("/etc/hosts");
+  });
+
+  test("resolves parent traversal normally", () => {
+    const resolved = resolvePath("/workspace/project", "../other/file.ts");
+    expect(resolved).toBe(resolve("/workspace/project", "../other/file.ts"));
+  });
+});

packages/workflow-agent-builtin/__tests__/prompt.test.ts

@@ -0,0 +1,236 @@
+import { describe, expect, test } from "bun:test";
+
+import type { AgentContext } from "@uncaged/workflow-agent-kit";
+
+import { buildBuiltinMessages } from "../src/prompt.js";
+
+function minimalContext(overrides: Partial<AgentContext> = {}): AgentContext {
+  return {
+    threadId: "00000000000000000000000000" as AgentContext["threadId"],
+    role: "developer",
+    store: {} as AgentContext["store"],
+    workflow: {
+      name: "test",
+      description: "test workflow",
+      roles: {
+        developer: {
+          description: "Developer role",
+          goal: "Ship the fix",
+          capabilities: ["file-edit"],
+          procedure: "Edit files",
+          output: "A patch",
+          frontmatter: "schema-hash",
+        },
+      },
+      conditions: {},
+      graph: {},
+    },
+    start: { workflow: "wf-hash", prompt: "Fix the bug" },
+    steps: [],
+    outputFormatInstruction: "---\nstatus: done\n---",
+    edgePrompt: "Implement the fix described in the plan.",
+    isFirstVisit: true,
+    ...overrides,
+  };
+}
+
+describe("buildBuiltinMessages", () => {
+  test("system includes output format and role goal", () => {
+    const messages = buildBuiltinMessages(minimalContext());
+    const system = messages[0];
+    expect(system?.role).toBe("system");
+    if (system?.role === "system") {
+      expect(system.content).toContain("status: done");
+      expect(system.content).toContain("## Goal");
+      expect(system.content).toContain("Ship the fix");
+    }
+  });
+
+  test("first visit produces system + single user message with edge prompt", () => {
+    const messages = buildBuiltinMessages(minimalContext());
+    expect(messages).toHaveLength(2);
+    expect(messages[1]?.role).toBe("user");
+    if (messages[1]?.role === "user") {
+      expect(messages[1].content).toContain("Implement the fix");
+      expect(messages[1].content).not.toContain("## What Happened Since Your Last Turn");
+    }
+  });
+
+  test("first visit with prior steps includes inter-step summary in final user message", () => {
+    const messages = buildBuiltinMessages(
+      minimalContext({
+        steps: [
+          {
+            role: "planner",
+            output: { plan: "step 1" },
+            agent: "uwf-builtin",
+            detail: "detail-hash",
+            edgePrompt: "Create a plan.",
+          },
+        ],
+      }),
+    );
+    expect(messages).toHaveLength(2);
+    const finalUser = messages[1];
+    if (finalUser?.role === "user") {
+      expect(finalUser.content).toContain("Implement the fix");
+      expect(finalUser.content).toContain("## What Happened Since Your Last Turn");
+      expect(finalUser.content).toContain("planner");
+    }
+  });
+
+  test("re-entry reconstructs prior user/assistant turns plus current user message", () => {
+    const messages = buildBuiltinMessages(
+      minimalContext({
+        isFirstVisit: false,
+        edgePrompt: "Fix the reviewer's feedback.",
+        steps: [
+          {
+            role: "developer",
+            output: { summary: "Initial fix" },
+            agent: "uwf-builtin",
+            detail: "detail-1",
+            edgePrompt: "Implement the fix.",
+          },
+          {
+            role: "reviewer",
+            output: { approved: false, comments: "Missing tests" },
+            agent: "uwf-builtin",
+            detail: "detail-2",
+            edgePrompt: "Review the implementation.",
+          },
+        ],
+      }),
+    );
+
+    expect(messages).toHaveLength(4);
+    expect(messages[0]?.role).toBe("system");
+    expect(messages[1]?.role).toBe("user");
+    expect(messages[2]?.role).toBe("assistant");
+    expect(messages[3]?.role).toBe("user");
+
+    if (messages[1]?.role === "user") {
+      expect(messages[1].content).toBe("Implement the fix.");
+    }
+    if (messages[2]?.role === "assistant") {
+      expect(messages[2].content).toBe(JSON.stringify({ summary: "Initial fix" }));
+    }
+    if (messages[3]?.role === "user") {
+      expect(messages[3].content).toContain("Fix the reviewer's feedback.");
+      expect(messages[3].content).toContain("## What Happened Since Your Last Turn");
+      expect(messages[3].content).toContain("reviewer");
+      expect(messages[3].content).toContain("Missing tests");
+    }
+  });
+
+  test("prefix is stable across re-entry for LLM cache hits", () => {
+    const firstVisitMessages = buildBuiltinMessages(
+      minimalContext({
+        edgePrompt: "Implement the fix.",
+        steps: [],
+      }),
+    );
+
+    const reEntryMessages = buildBuiltinMessages(
+      minimalContext({
+        isFirstVisit: false,
+        edgePrompt: "Fix the reviewer's feedback.",
+        steps: [
+          {
+            role: "developer",
+            output: { summary: "Initial fix" },
+            agent: "uwf-builtin",
+            detail: "detail-1",
+            edgePrompt: "Implement the fix.",
+          },
+          {
+            role: "reviewer",
+            output: { approved: false },
+            agent: "uwf-builtin",
+            detail: "detail-2",
+            edgePrompt: "Review the code.",
+          },
+        ],
+      }),
+    );
+
+    expect(reEntryMessages[0]).toEqual(firstVisitMessages[0]);
+    expect(reEntryMessages[1]).toEqual(firstVisitMessages[1]);
+    expect(reEntryMessages[2]?.role).toBe("assistant");
+    if (reEntryMessages[2]?.role === "assistant") {
+      expect(reEntryMessages[2].content).toBe(JSON.stringify({ summary: "Initial fix" }));
+    }
+    expect(reEntryMessages[3]?.role).toBe("user");
+    if (reEntryMessages[3]?.role === "user") {
+      expect(reEntryMessages[3].content).toContain("Fix the reviewer's feedback.");
+    }
+  });
+
+  test("multiple prior visits emit one user/assistant pair per visit", () => {
+    const messages = buildBuiltinMessages(
+      minimalContext({
+        isFirstVisit: false,
+        edgePrompt: "Third round fix.",
+        steps: [
+          {
+            role: "developer",
+            output: { round: 1 },
+            agent: "uwf-builtin",
+            detail: "d1",
+            edgePrompt: "First attempt.",
+          },
+          {
+            role: "reviewer",
+            output: { approved: false },
+            agent: "uwf-builtin",
+            detail: "d2",
+            edgePrompt: "Review round 1.",
+          },
+          {
+            role: "developer",
+            output: { round: 2 },
+            agent: "uwf-builtin",
+            detail: "d3",
+            edgePrompt: "Second attempt.",
+          },
+          {
+            role: "reviewer",
+            output: { approved: false },
+            agent: "uwf-builtin",
+            detail: "d4",
+            edgePrompt: "Review round 2.",
+          },
+        ],
+      }),
+    );
+
+    expect(messages).toHaveLength(6);
+    expect(messages.map((m) => m.role)).toEqual([
+      "system",
+      "user",
+      "assistant",
+      "user",
+      "assistant",
+      "user",
+    ]);
+
+    if (messages[1]?.role === "user") {
+      expect(messages[1].content).toBe("First attempt.");
+    }
+    if (messages[2]?.role === "assistant") {
+      expect(messages[2].content).toBe(JSON.stringify({ round: 1 }));
+    }
+    if (messages[3]?.role === "user") {
+      expect(messages[3].content).toContain("Second attempt.");
+      expect(messages[3].content).toContain("reviewer");
+    }
+    if (messages[4]?.role === "assistant") {
+      expect(messages[4].content).toBe(JSON.stringify({ round: 2 }));
+    }
+    if (messages[5]?.role === "user") {
+      expect(messages[5].content).toContain("Third round fix.");
+      expect(messages[5].content).toContain("### Step 4: reviewer");
+      expect(messages[5].content).toContain('"approved":false');
+    }
+  });
+});

packages/workflow-agent-builtin/package.json

@@ -0,0 +1,34 @@
+{
+  "name": "@uncaged/workflow-agent-builtin",
+  "version": "0.5.0",
+  "files": [
+    "src",
+    "dist",
+    "package.json"
+  ],
+  "type": "module",
+  "bin": {
+    "uwf-builtin": "./src/cli.ts"
+  },
+  "exports": {
+    ".": {
+      "bun": "./src/index.ts",
+      "types": "./dist/index.d.ts",
+      "import": "./dist/index.js"
+    }
+  },
+  "scripts": {
+    "test": "bun test"
+  },
+  "dependencies": {
+    "@uncaged/json-cas": "^0.4.0",
+    "@uncaged/workflow-agent-kit": "workspace:^",
+    "@uncaged/workflow-util": "workspace:^"
+  },
+  "devDependencies": {
+    "typescript": "^5.8.3"
+  },
+  "publishConfig": {
+    "access": "public"
+  }
+}

packages/workflow-agent-builtin/src/agent.ts

@@ -0,0 +1,158 @@
+import type { Store } from "@uncaged/json-cas";
+import {
+  type AgentContext,
+  type AgentRunResult,
+  createAgent,
+  loadWorkflowConfig,
+  resolveModel,
+  resolveStorageRoot,
+} from "@uncaged/workflow-agent-kit";
+import { createLogger, generateUlid } from "@uncaged/workflow-util";
+
+import { storeBuiltinDetail } from "./detail.js";
+import type { ChatMessage } from "./llm/index.js";
+import { BUILTIN_CONTINUE_MAX_TURNS, BUILTIN_MAX_TURNS, runBuiltinLoop } from "./loop.js";
+import { buildBuiltinMessages } from "./prompt.js";
+import { initSessionDir } from "./session.js";
+
+const log = createLogger({ sink: { kind: "stderr" } });
+
+const FRONTMATTER_FENCE = "---";
+
+/**
+ * Strip any text before the first `---` fence.
+ * LLMs sometimes emit preamble text before the frontmatter block.
+ */
+function stripPreamble(text: string): string {
+  if (text.startsWith(FRONTMATTER_FENCE)) {
+    return text;
+  }
+  const idx = text.indexOf(`\n${FRONTMATTER_FENCE}\n`);
+  if (idx !== -1) {
+    log("6GWRP3QX", `stripped ${idx + 1} chars of preamble before frontmatter`);
+    return text.slice(idx + 1);
+  }
+  return text;
+}
+
+type SessionRecord = {
+  sessionId: string;
+  model: string;
+  startedAtMs: number;
+  messages: ChatMessage[];
+};
+
+const sessions = new Map<string, SessionRecord>();
+
+function getSession(sessionId: string): SessionRecord {
+  const session = sessions.get(sessionId);
+  if (session === undefined) {
+    throw new Error(`builtin session not found: ${sessionId}`);
+  }
+  return session;
+}
+
+function buildToolContext(storageRoot: string): { cwd: string; storageRoot: string } {
+  return {
+    cwd: process.cwd(),
+    storageRoot,
+  };
+}
+
+async function runBuiltinWithMessages(
+  storageRoot: string,
+  provider: ReturnType<typeof resolveModel>,
+  messages: ChatMessage[],
+  session: SessionRecord,
+  store: Store,
+  maxTurns: number,
+  noTools: boolean,
+): Promise<AgentRunResult> {
+  const loopResult = await runBuiltinLoop({
+    provider,
+    messages,
+    toolCtx: buildToolContext(storageRoot),
+    maxTurns,
+    storageRoot,
+    sessionId: session.sessionId,
+    noTools,
+  });
+
+  session.messages = loopResult.messages;
+
+  if (loopResult.turnCount === 0) {
+    log("5RWTK9NB", "no turns produced, returning empty output");
+    return { output: "", detailHash: "", sessionId: session.sessionId };
+  }
+
+  // Read jsonl → persist turns to CAS → store detail
+  const { detailHash } = await storeBuiltinDetail(
+    store,
+    storageRoot,
+    session.sessionId,
+    session.model,
+    session.startedAtMs,
+  );
+
+  return { output: stripPreamble(loopResult.finalText), detailHash, sessionId: session.sessionId };
+}
+
+async function runBuiltin(ctx: AgentContext): Promise<AgentRunResult> {
+  const storageRoot = resolveStorageRoot();
+  const config = await loadWorkflowConfig(storageRoot);
+  const provider = resolveModel(config, config.defaultModel);
+
+  const sessionId = generateUlid(Date.now());
+  await initSessionDir(storageRoot);
+  const messages = buildBuiltinMessages(ctx);
+
+  const session: SessionRecord = {
+    sessionId,
+    model: provider.model,
+    startedAtMs: Date.now(),
+    messages,
+  };
+  sessions.set(sessionId, session);
+
+  return runBuiltinWithMessages(
+    storageRoot,
+    provider,
+    messages,
+    session,
+    ctx.store,
+    BUILTIN_MAX_TURNS,
+    false,
+  );
+}
+
+async function continueBuiltin(
+  sessionId: string,
+  message: string,
+  store: Store,
+): Promise<AgentRunResult> {
+  const session = getSession(sessionId);
+  const storageRoot = resolveStorageRoot();
+  const config = await loadWorkflowConfig(storageRoot);
+  const provider = resolveModel(config, config.defaultModel);
+
+  const messages: ChatMessage[] = [...session.messages, { role: "user", content: message }];
+
+  return runBuiltinWithMessages(
+    storageRoot,
+    provider,
+    messages,
+    session,
+    store,
+    BUILTIN_CONTINUE_MAX_TURNS,
+    true,
+  );
+}
+
+/** Agent CLI factory: built-in LLM loop with file/shell tools. */
+export function createBuiltinAgent(): () => Promise<void> {
+  return createAgent({
+    name: "builtin",
+    run: runBuiltin,
+    continue: continueBuiltin,
+  });
+}

packages/workflow-agent-builtin/src/cli.ts

@@ -0,0 +1,6 @@
+#!/usr/bin/env bun
+
+import { createBuiltinAgent } from "./agent.js";
+
+const main = createBuiltinAgent();
+void main();

packages/workflow-agent-builtin/src/detail.ts

@@ -0,0 +1,49 @@
+import { bootstrap, putSchema, type Store } from "@uncaged/json-cas";
+
+import { BUILTIN_DETAIL_SCHEMA, BUILTIN_TURN_SCHEMA } from "./schemas.js";
+import { readSessionTurns } from "./session.js";
+import type { BuiltinDetailPayload } from "./types.js";
+
+type BuiltinSchemaHashes = {
+  turn: string;
+  detail: string;
+};
+
+export async function registerBuiltinSchemas(store: Store): Promise<BuiltinSchemaHashes> {
+  await bootstrap(store);
+  const [turn, detail] = await Promise.all([
+    putSchema(store, BUILTIN_TURN_SCHEMA),
+    putSchema(store, BUILTIN_DETAIL_SCHEMA),
+  ]);
+  return { turn, detail };
+}
+
+/** Read session jsonl, persist each turn to CAS, return detail hash. */
+export async function storeBuiltinDetail(
+  store: Store,
+  storageRoot: string,
+  sessionId: string,
+  model: string,
+  startedAtMs: number,
+  nowMs: number = Date.now(),
+): Promise<{ detailHash: string; turnCount: number }> {
+  const schemas = await registerBuiltinSchemas(store);
+  const turns = await readSessionTurns(storageRoot, sessionId);
+
+  const turnHashes: string[] = [];
+  for (const turn of turns) {
+    const hash = await store.put(schemas.turn, turn);
+    turnHashes.push(hash);
+  }
+
+  const duration = Math.max(0, nowMs - startedAtMs);
+  const detail: BuiltinDetailPayload = {
+    sessionId,
+    model,
+    duration,
+    turnCount: turnHashes.length,
+    turns: turnHashes,
+  };
+  const detailHash = await store.put(schemas.detail, detail);
+  return { detailHash, turnCount: turnHashes.length };
+}

packages/workflow-agent-builtin/src/index.ts

@@ -0,0 +1,16 @@
+export { createBuiltinAgent } from "./agent.js";
+export { registerBuiltinSchemas, storeBuiltinDetail } from "./detail.js";
+export type { ChatMessage, LlmAssistantResponse, LlmToolCall } from "./llm/index.js";
+export { chatCompletionWithTools } from "./llm/index.js";
+export { BUILTIN_CONTINUE_MAX_TURNS, BUILTIN_MAX_TURNS, runBuiltinLoop } from "./loop.js";
+export { buildBuiltinMessages } from "./prompt.js";
+export { appendSessionTurn, initSessionDir, readSessionTurns, removeSession } from "./session.js";
+export type { BuiltinTool, ToolContext } from "./tools/index.js";
+export { executeBuiltinTool, getBuiltinTools } from "./tools/index.js";
+export type {
+  BuiltinDetailPayload,
+  BuiltinLoopTurn,
+  BuiltinToolCallRecord,
+  BuiltinToolResultRecord,
+  BuiltinTurnPayload,
+} from "./types.js";

packages/workflow-agent-builtin/src/llm/index.ts

@@ -0,0 +1,7 @@
+export { chatCompletionWithTools } from "./llm.js";
+export type {
+  ChatMessage,
+  LlmAssistantResponse,
+  LlmToolCall,
+  OpenAiToolDefinition,
+} from "./types.js";

packages/workflow-agent-builtin/src/llm/llm.ts

@@ -0,0 +1,139 @@
+import type { ResolvedLlmProvider } from "@uncaged/workflow-agent-kit";
+
+import type {
+  ChatMessage,
+  LlmAssistantResponse,
+  LlmToolCall,
+  OpenAiToolDefinition,
+} from "./types.js";
+
+function isRecord(value: unknown): value is Record<string, unknown> {
+  return typeof value === "object" && value !== null && !Array.isArray(value);
+}
+
+function chatUrl(baseUrl: string): string {
+  const trimmed = baseUrl.replace(/\/+$/, "");
+  return `${trimmed}/chat/completions`;
+}
+
+function parseToolCalls(raw: unknown): LlmToolCall[] | null {
+  if (!Array.isArray(raw) || raw.length === 0) {
+    return null;
+  }
+  const calls: LlmToolCall[] = [];
+  for (const entry of raw) {
+    if (!isRecord(entry)) {
+      continue;
+    }
+    const id = entry.id;
+    const fn = entry.function;
+    if (typeof id !== "string" || !isRecord(fn)) {
+      continue;
+    }
+    const name = fn.name;
+    const args = fn.arguments;
+    if (typeof name !== "string" || typeof args !== "string") {
+      continue;
+    }
+    calls.push({ id, name, arguments: args });
+  }
+  return calls.length > 0 ? calls : null;
+}
+
+function parseAssistantMessage(parsed: unknown): LlmAssistantResponse {
+  if (!isRecord(parsed)) {
+    throw new Error("LLM response is not an object");
+  }
+  const choices = parsed.choices;
+  if (!Array.isArray(choices) || choices.length === 0) {
+    throw new Error("LLM response has no choices");
+  }
+  const c0 = choices[0];
+  if (!isRecord(c0)) {
+    throw new Error("LLM choice is not an object");
+  }
+  const messageObj = c0.message;
+  if (!isRecord(messageObj)) {
+    throw new Error("LLM message is not an object");
+  }
+  const contentRaw = messageObj.content;
+  const content =
+    typeof contentRaw === "string"
+      ? contentRaw
+      : contentRaw === null || contentRaw === undefined
+        ? null
+        : null;
+  const toolCalls = parseToolCalls(messageObj.tool_calls);
+  return { content, toolCalls };
+}
+
+function serializeMessage(message: ChatMessage): Record<string, unknown> {
+  if (message.role === "tool") {
+    return {
+      role: "tool",
+      tool_call_id: message.tool_call_id,
+      content: message.content,
+    };
+  }
+  if (message.role === "assistant") {
+    const base: Record<string, unknown> = {
+      role: "assistant",
+      content: message.content,
+    };
+    if (message.tool_calls !== null && message.tool_calls.length > 0) {
+      base.tool_calls = message.tool_calls.map((call) => ({
+        id: call.id,
+        type: "function",
+        function: { name: call.name, arguments: call.arguments },
+      }));
+    }
+    return base;
+  }
+  return { role: message.role, content: message.content };
+}
+
+/** OpenAI-compatible chat completion with tool calling (non-streaming). */
+export async function chatCompletionWithTools(
+  provider: ResolvedLlmProvider,
+  messages: ChatMessage[],
+  tools: OpenAiToolDefinition[] | null,
+): Promise<LlmAssistantResponse> {
+  const body: Record<string, unknown> = {
+    model: provider.model,
+    messages: messages.map(serializeMessage),
+  };
+  if (tools !== null && tools.length > 0) {
+    body.tools = tools;
+    body.tool_choice = "auto";
+  }
+
+  let response: Response;
+  try {
+    response = await fetch(chatUrl(provider.baseUrl), {
+      method: "POST",
+      headers: {
+        Authorization: `Bearer ${provider.apiKey}`,
+        "Content-Type": "application/json",
+      },
+      body: JSON.stringify(body),
+    });
+  } catch (cause) {
+    const message = cause instanceof Error ? cause.message : String(cause);
+    throw new Error(`LLM network error: ${message}`);
+  }
+
+  const responseText = await response.text();
+  if (!response.ok) {
+    throw new Error(`LLM HTTP ${response.status}: ${responseText.slice(0, 2000)}`);
+  }
+
+  let parsed: unknown;
+  try {
+    parsed = JSON.parse(responseText) as unknown;
+  } catch (cause) {
+    const message = cause instanceof Error ? cause.message : String(cause);
+    throw new Error(`LLM invalid JSON response: ${message}`);
+  }
+
+  return parseAssistantMessage(parsed);
+}

packages/workflow-agent-builtin/src/llm/types.ts

@@ -0,0 +1,29 @@
+export type LlmToolCall = {
+  id: string;
+  name: string;
+  arguments: string;
+};
+
+export type LlmAssistantResponse = {
+  content: string | null;
+  toolCalls: LlmToolCall[] | null;
+};
+
+export type ChatMessage =
+  | { role: "system"; content: string }
+  | { role: "user"; content: string }
+  | {
+      role: "assistant";
+      content: string | null;
+      tool_calls: LlmToolCall[] | null;
+    }
+  | { role: "tool"; tool_call_id: string; content: string };
+
+export type OpenAiToolDefinition = {
+  type: "function";
+  function: {
+    name: string;
+    description: string;
+    parameters: Record<string, unknown>;
+  };
+};

packages/workflow-agent-builtin/src/loop.ts

@@ -0,0 +1,303 @@
+import type { ResolvedLlmProvider } from "@uncaged/workflow-agent-kit";
+import { createLogger } from "@uncaged/workflow-util";
+
+import {
+  type ChatMessage,
+  chatCompletionWithTools,
+  type LlmToolCall,
+  type OpenAiToolDefinition,
+} from "./llm/index.js";
+import { appendSessionTurn } from "./session.js";
+import {
+  builtinToolsToOpenAi,
+  executeBuiltinTool,
+  getBuiltinTools,
+  type ToolContext,
+} from "./tools/index.js";
+import type { BuiltinToolCall, BuiltinTurnPayload } from "./types.js";
+
+const log = createLogger({ sink: { kind: "stderr" } });
+
+export const BUILTIN_MAX_TURNS = 30;
+export const BUILTIN_CONTINUE_MAX_TURNS = 5;
+
+export type RunBuiltinLoopOptions = {
+  provider: ResolvedLlmProvider;
+  messages: ChatMessage[];
+  toolCtx: ToolContext;
+  maxTurns: number;
+  storageRoot: string;
+  sessionId: string;
+  /** When true, do not provide tools — force LLM to emit text only. */
+  noTools: boolean;
+};
+
+export type RunBuiltinLoopResult = {
+  finalText: string;
+  messages: ChatMessage[];
+  turnCount: number;
+};
+
+function mapToolCallsForPayload(calls: LlmToolCall[]): BuiltinToolCall[] {
+  return calls.map((call) => ({
+    name: call.name,
+    args: call.arguments,
+  }));
+}
+
+async function appendTurn(
+  storageRoot: string,
+  sessionId: string,
+  payload: BuiltinTurnPayload,
+): Promise<void> {
+  await appendSessionTurn(storageRoot, sessionId, payload);
+}
+
+export async function executeTurnTools(
+  calls: Array<{ id: string; name: string; arguments: string }>,
+  toolCtx: ToolContext,
+  messages: ChatMessage[],
+  storageRoot: string,
+  sessionId: string,
+): Promise<number> {
+  let turnCount = 0;
+  for (const call of calls) {
+    const result = await executeBuiltinTool(call.name, call.arguments, toolCtx);
+    messages.push({ role: "tool", tool_call_id: call.id, content: result });
+    await appendTurn(storageRoot, sessionId, {
+      role: "tool",
+      content: result,
+      toolCalls: null,
+      reasoning: null,
+    });
+    turnCount += 1;
+  }
+  return turnCount;
+}
+
+export type ShouldNudgeOptions = {
+  noTools: boolean;
+  text: string;
+  turn: number;
+  maxTurns: number;
+};
+
+const MAX_NUDGES = 3;
+const DEADLINE_WARNING_TURNS = 3;
+
+export function shouldInjectDeadlineWarning(
+  turn: number,
+  maxTurns: number,
+  alreadyWarned: boolean,
+  noTools: boolean,
+): boolean {
+  const turnsRemaining = maxTurns - turn;
+  return (
+    !noTools && !alreadyWarned && turnsRemaining > 0 && turnsRemaining <= DEADLINE_WARNING_TURNS
+  );
+}
+
+export function shouldProcessToolCalls(toolCalls: LlmToolCall[] | null, noTools: boolean): boolean {
+  return !noTools && toolCalls !== null && toolCalls.length > 0;
+}
+
+export function extractFinalText(messages: ChatMessage[]): string {
+  for (let i = messages.length - 1; i >= 0; i--) {
+    const msg = messages[i];
+    if (
+      msg !== undefined &&
+      msg.role === "assistant" &&
+      msg.content !== null &&
+      msg.content.trim() !== ""
+    ) {
+      return msg.content;
+    }
+  }
+  return "";
+}
+
+function injectDeadlineWarning(messages: ChatMessage[], turnsRemaining: number): void {
+  log("4NRXW6KT", `${turnsRemaining} turns remaining, injecting deadline warning`);
+  messages.push({
+    role: "user",
+    content:
+      `⚠️ You have ${turnsRemaining} turns remaining. ` +
+      "Wrap up your work and output the YAML frontmatter starting with `---`. " +
+      "If you cannot finish in time, output frontmatter with `status: failed` and describe what remains.",
+  });
+}
+
+type HandleTextOnlyTurnResult = {
+  shouldBreak: boolean;
+  finalText: string;
+  turnCount: number;
+  nudgeCount: number;
+  turnAdjustment: number;
+};
+
+async function handleTextOnlyTurn(
+  text: string,
+  messages: ChatMessage[],
+  storageRoot: string,
+  sessionId: string,
+  noTools: boolean,
+  turn: number,
+  maxTurns: number,
+  currentNudgeCount: number,
+): Promise<HandleTextOnlyTurnResult> {
+  await appendTurn(storageRoot, sessionId, {
+    role: "assistant",
+    content: text,
+    toolCalls: null,
+    reasoning: null,
+  });
+  const turnCount = 1;
+  let nudgeCount = currentNudgeCount;
+  let turnAdjustment = 0;
+
+  if (shouldNudge({ noTools, text, turn, maxTurns })) {
+    nudgeCount += 1;
+    log("7FXQM2KN", `text-only turn without frontmatter, nudge ${nudgeCount}/${MAX_NUDGES}`);
+    const nudge =
+      "You stopped calling tools but your response does not start with the required `---` YAML frontmatter. " +
+      "Either continue using tools to complete your work, or output your final response starting with `---`.";
+    messages.push({ role: "user", content: nudge });
+    // Nudge doesn't consume turn budget (up to MAX_NUDGES)
+    if (nudgeCount <= MAX_NUDGES) {
+      turnAdjustment = -1;
+    }
+    return { shouldBreak: false, finalText: "", turnCount, nudgeCount, turnAdjustment };
+  }
+
+  return { shouldBreak: true, finalText: text, turnCount, nudgeCount, turnAdjustment };
+}
+
+async function handleToolCallTurn(
+  content: string,
+  toolCalls: LlmToolCall[],
+  messages: ChatMessage[],
+  storageRoot: string,
+  sessionId: string,
+  toolCtx: ToolContext,
+): Promise<number> {
+  await appendTurn(storageRoot, sessionId, {
+    role: "assistant",
+    content,
+    toolCalls: mapToolCallsForPayload(toolCalls),
+    reasoning: null,
+  });
+  let turnCount = 1;
+
+  // Execute tools
+  turnCount += await executeTurnTools(toolCalls, toolCtx, messages, storageRoot, sessionId);
+
+  return turnCount;
+}
+
+export function shouldNudge({ noTools, text, turn, maxTurns }: ShouldNudgeOptions): boolean {
+  return !noTools && !text.trimStart().startsWith("---") && turn < maxTurns - 1;
+}
+
+type ProcessLoopIterationResult = {
+  shouldBreak: boolean;
+  finalText: string;
+  turnCount: number;
+  nudgeCount: number;
+  turnAdjustment: number;
+};
+
+async function processLoopIteration(
+  options: RunBuiltinLoopOptions,
+  messages: ChatMessage[],
+  openAiTools: OpenAiToolDefinition[],
+  turn: number,
+  nudgeCount: number,
+): Promise<ProcessLoopIterationResult> {
+  const response = await chatCompletionWithTools(
+    options.provider,
+    messages,
+    openAiTools.length > 0 ? openAiTools : null,
+  );
+
+  // When noTools is set, ignore any tool_calls the LLM might still return
+  const effectiveToolCalls = options.noTools ? null : (response.toolCalls ?? null);
+
+  const assistantMessage: ChatMessage = {
+    role: "assistant",
+    content: response.content,
+    tool_calls: effectiveToolCalls,
+  };
+  messages.push(assistantMessage);
+
+  if (!shouldProcessToolCalls(effectiveToolCalls, options.noTools)) {
+    const text = response.content ?? "";
+    const result = await handleTextOnlyTurn(
+      text,
+      messages,
+      options.storageRoot,
+      options.sessionId,
+      options.noTools,
+      turn,
+      options.maxTurns,
+      nudgeCount,
+    );
+    return result;
+  }
+
+  // At this point, effectiveToolCalls is guaranteed to be non-null and non-empty
+  const turnCount = await handleToolCallTurn(
+    response.content ?? "",
+    effectiveToolCalls as LlmToolCall[],
+    messages,
+    options.storageRoot,
+    options.sessionId,
+    options.toolCtx,
+  );
+
+  return {
+    shouldBreak: false,
+    finalText: "",
+    turnCount,
+    nudgeCount,
+    turnAdjustment: 0,
+  };
+}
+
+/** Agent run loop: LLM ↔ tools until no tool_calls or maxTurns. */
+export async function runBuiltinLoop(
+  options: RunBuiltinLoopOptions,
+): Promise<RunBuiltinLoopResult> {
+  const messages = [...options.messages];
+  const openAiTools = options.noTools ? [] : builtinToolsToOpenAi(getBuiltinTools());
+  let finalText = "";
+  let turnCount = 0;
+  let nudgeCount = 0;
+  let deadlineWarned = false;
+
+  for (let turn = 0; turn < options.maxTurns; turn++) {
+    log("8K2M4N7P", `builtin loop turn ${turn + 1}/${options.maxTurns}`);
+
+    // Warn agent when approaching turn limit
+    if (shouldInjectDeadlineWarning(turn, options.maxTurns, deadlineWarned, options.noTools)) {
+      deadlineWarned = true;
+      const turnsRemaining = options.maxTurns - turn;
+      injectDeadlineWarning(messages, turnsRemaining);
+    }
+
+    const result = await processLoopIteration(options, messages, openAiTools, turn, nudgeCount);
+    turnCount += result.turnCount;
+    nudgeCount = result.nudgeCount;
+    turn += result.turnAdjustment;
+
+    if (result.shouldBreak) {
+      finalText = result.finalText;
+      break;
+    }
+  }
+
+  if (finalText === "") {
+    finalText = extractFinalText(messages);
+  }
+
+  return { finalText, messages, turnCount };
+}

packages/workflow-agent-builtin/src/prompt.ts

@@ -0,0 +1,115 @@
+import { type AgentContext, buildRolePrompt } from "@uncaged/workflow-agent-kit";
+
+import type { ChatMessage } from "./llm/index.js";
+
+type StepContext = AgentContext["steps"][number];
+
+function formatStep(step: StepContext, stepNumber: number): string {
+  return [
+    `### Step ${stepNumber}: ${step.role}`,
+    `Output: ${JSON.stringify(step.output)}`,
+    `Agent: ${step.agent}`,
+  ].join("\n");
+}
+
+function buildStepsSummary(steps: StepContext[], fromIndex: number, toIndex: number): string {
+  if (fromIndex >= toIndex) {
+    return "";
+  }
+
+  const lines: string[] = ["## What Happened Since Your Last Turn"];
+  for (let i = fromIndex; i < toIndex; i++) {
+    const step = steps[i];
+    if (step === undefined) {
+      continue;
+    }
+    lines.push("");
+    lines.push(formatStep(step, i + 1));
+  }
+  return lines.join("\n");
+}
+
+function buildUserTurnContent(edgePrompt: string, summary: string): string {
+  const parts: string[] = [];
+  if (edgePrompt !== "") {
+    parts.push(edgePrompt);
+  }
+  if (summary !== "") {
+    if (parts.length > 0) {
+      parts.push("");
+    }
+    parts.push(summary);
+  }
+  return parts.join("\n");
+}
+
+/**
+ * Reconstruct multi-turn chat messages from thread history for cache-friendly session resume.
+ *
+ * - system: role prompt + output format (stable prefix)
+ * - For each prior visit of this role: user (edgePrompt + inter-step summary) + assistant (output JSON)
+ * - Final user: current edgePrompt + summary since last visit of this role
+ */
+export function buildBuiltinMessages(ctx: AgentContext): ChatMessage[] {
+  const roleDef = ctx.workflow.roles[ctx.role];
+  const rolePrompt = roleDef !== undefined ? buildRolePrompt(roleDef) : "";
+  const systemParts: string[] = [];
+  if (ctx.outputFormatInstruction !== "") {
+    systemParts.push(ctx.outputFormatInstruction, "");
+  }
+  systemParts.push(rolePrompt);
+
+  systemParts.push(
+    "",
+    "## Workflow",
+    "",
+    `Your working directory is: ${process.cwd()}`,
+    "",
+    "You have tools available (read_file, write_file, run_command). " +
+      "Use them to complete your task — read files, run commands, make changes as needed. " +
+      "Your task is described in the user message below — do NOT use uwf or workflow CLI commands to discover your task. " +
+      "When you are done, output your final response with the YAML frontmatter block as specified above. " +
+      "Do NOT output the frontmatter until you have completed all necessary work. " +
+      "If you are running low on turns and cannot finish, output the frontmatter with `status: failed` and explain what remains in the body. " +
+      "CRITICAL: Your final output MUST start with the `---` fence on the very first line — " +
+      "no preamble text, no explanation before it. The parser requires `---` at position 0.",
+  );
+
+  const messages: ChatMessage[] = [{ role: "system", content: systemParts.join("\n") }];
+
+  const roleVisitIndices: number[] = [];
+  for (let i = 0; i < ctx.steps.length; i++) {
+    const step = ctx.steps[i];
+    if (step !== undefined && step.role === ctx.role) {
+      roleVisitIndices.push(i);
+    }
+  }
+
+  let prevVisitIndex = -1;
+  for (const visitIndex of roleVisitIndices) {
+    const visitStep = ctx.steps[visitIndex];
+    if (visitStep === undefined) {
+      continue;
+    }
+
+    const summary = buildStepsSummary(ctx.steps, prevVisitIndex + 1, visitIndex);
+    messages.push({
+      role: "user",
+      content: buildUserTurnContent(visitStep.edgePrompt, summary),
+    });
+    messages.push({
+      role: "assistant",
+      content: JSON.stringify(visitStep.output),
+      tool_calls: null,
+    });
+    prevVisitIndex = visitIndex;
+  }
+
+  const finalSummary = buildStepsSummary(ctx.steps, prevVisitIndex + 1, ctx.steps.length);
+  messages.push({
+    role: "user",
+    content: buildUserTurnContent(ctx.edgePrompt, finalSummary),
+  });
+
+  return messages;
+}

packages/workflow-agent-builtin/src/schemas.ts

@@ -0,0 +1,45 @@
+import type { JSONSchema } from "@uncaged/json-cas";
+
+const BUILTIN_TOOL_CALL_SCHEMA: JSONSchema = {
+  type: "object",
+  required: ["name", "args"],
+  properties: {
+    name: { type: "string" },
+    args: { type: "string" },
+  },
+  additionalProperties: false,
+};
+
+export const BUILTIN_TURN_SCHEMA: JSONSchema = {
+  title: "builtin-turn",
+  type: "object",
+  required: ["role", "content"],
+  properties: {
+    role: { type: "string", enum: ["assistant", "tool"] },
+    content: { type: "string" },
+    toolCalls: {
+      anyOf: [{ type: "array", items: BUILTIN_TOOL_CALL_SCHEMA }, { type: "null" }],
+    },
+    reasoning: {
+      anyOf: [{ type: "string" }, { type: "null" }],
+    },
+  },
+  additionalProperties: false,
+};
+
+export const BUILTIN_DETAIL_SCHEMA: JSONSchema = {
+  title: "builtin-detail",
+  type: "object",
+  required: ["sessionId", "model", "duration", "turnCount", "turns"],
+  properties: {
+    sessionId: { type: "string" },
+    model: { type: "string" },
+    duration: { type: "integer" },
+    turnCount: { type: "integer" },
+    turns: {
+      type: "array",
+      items: { type: "string", format: "cas_ref" },
+    },
+  },
+  additionalProperties: false,
+};

packages/workflow-agent-builtin/src/session.ts

@@ -0,0 +1,59 @@
+import { appendFile, mkdir, readFile, rm } from "node:fs/promises";
+import { join } from "node:path";
+
+import { createLogger } from "@uncaged/workflow-util";
+
+import type { BuiltinTurnPayload } from "./types.js";
+
+const log = createLogger({ sink: { kind: "stderr" } });
+
+function sessionsDir(storageRoot: string): string {
+  return join(storageRoot, "sessions");
+}
+
+function sessionFile(storageRoot: string, sessionId: string): string {
+  return join(sessionsDir(storageRoot), `${sessionId}.jsonl`);
+}
+
+/** Ensure sessions directory exists. */
+export async function initSessionDir(storageRoot: string): Promise<void> {
+  await mkdir(sessionsDir(storageRoot), { recursive: true });
+}
+
+/** Append a turn to the session jsonl file. */
+export async function appendSessionTurn(
+  storageRoot: string,
+  sessionId: string,
+  turn: BuiltinTurnPayload,
+): Promise<void> {
+  const line = `${JSON.stringify(turn)}\n`;
+  await appendFile(sessionFile(storageRoot, sessionId), line, "utf-8");
+  log("3XQVN8KR", `session ${sessionId} appended ${turn.role} turn`);
+}
+
+/** Read all turns from session jsonl. Returns empty array if file does not exist. */
+export async function readSessionTurns(
+  storageRoot: string,
+  sessionId: string,
+): Promise<BuiltinTurnPayload[]> {
+  try {
+    const content = await readFile(sessionFile(storageRoot, sessionId), "utf-8");
+    const lines = content
+      .trim()
+      .split("\n")
+      .filter((l) => l.length > 0);
+    return lines.map((l) => JSON.parse(l) as BuiltinTurnPayload);
+  } catch {
+    return [];
+  }
+}
+
+/** Remove session jsonl file (called after detail is persisted to step CAS). */
+export async function removeSession(storageRoot: string, sessionId: string): Promise<void> {
+  try {
+    await rm(sessionFile(storageRoot, sessionId));
+    log("7FWDP2MJ", `session ${sessionId} removed`);
+  } catch {
+    // already gone — fine
+  }
+}

packages/workflow-agent-builtin/src/tools/index.ts

@@ -0,0 +1,44 @@
+import type { OpenAiToolDefinition } from "../llm/index.js";
+
+import { readFileTool } from "./read-file.js";
+import { runCommandTool } from "./run-command.js";
+import type { BuiltinTool, ToolContext } from "./types.js";
+import { writeFileTool } from "./write-file.js";
+
+export { resolvePath } from "./path.js";
+export type { BuiltinTool, ToolContext } from "./types.js";
+
+const BUILTIN_TOOLS: BuiltinTool[] = [readFileTool, writeFileTool, runCommandTool];
+
+export function getBuiltinTools(): readonly BuiltinTool[] {
+  return BUILTIN_TOOLS;
+}
+
+export function builtinToolsToOpenAi(tools: readonly BuiltinTool[]): OpenAiToolDefinition[] {
+  return tools.map((tool) => ({
+    type: "function",
+    function: {
+      name: tool.name,
+      description: tool.description,
+      parameters: tool.parameters as Record<string, unknown>,
+    },
+  }));
+}
+
+export async function executeBuiltinTool(
+  name: string,
+  argsJson: string,
+  ctx: ToolContext,
+): Promise<string> {
+  const tool = BUILTIN_TOOLS.find((t) => t.name === name);
+  if (tool === undefined) {
+    return `Error: unknown tool ${name}`;
+  }
+  let args: unknown;
+  try {
+    args = JSON.parse(argsJson) as unknown;
+  } catch {
+    return "Error: tool arguments must be valid JSON";
+  }
+  return tool.execute(args, ctx);
+}

packages/workflow-agent-builtin/src/tools/path.ts

@@ -0,0 +1,6 @@
+import { resolve } from "node:path";
+
+/** Resolve a path relative to the working directory. */
+export function resolvePath(cwd: string, inputPath: string): string {
+  return resolve(cwd, inputPath);
+}

packages/workflow-agent-builtin/src/tools/read-file.ts

@@ -0,0 +1,41 @@
+import { readFile, stat } from "node:fs/promises";
+import { resolvePath } from "./path.js";
+import type { BuiltinTool } from "./types.js";
+
+const MAX_READ_BYTES = 512 * 1024;
+
+function isRecord(value: unknown): value is Record<string, unknown> {
+  return typeof value === "object" && value !== null && !Array.isArray(value);
+}
+
+export const readFileTool: BuiltinTool = {
+  name: "read_file",
+  description: "Read a UTF-8 text file from the workspace.",
+  parameters: {
+    type: "object",
+    required: ["path"],
+    properties: {
+      path: { type: "string", description: "Relative or absolute path within the workspace." },
+    },
+    additionalProperties: false,
+  },
+  execute: async (args, ctx) => {
+    if (!isRecord(args) || typeof args.path !== "string") {
+      return "Error: path must be a string";
+    }
+    const resolved = resolvePath(ctx.cwd, args.path);
+    try {
+      const info = await stat(resolved);
+      if (!info.isFile()) {
+        return "Error: not a file";
+      }
+      if (info.size > MAX_READ_BYTES) {
+        return `Error: file exceeds ${MAX_READ_BYTES} byte limit`;
+      }
+      return await readFile(resolved, "utf8");
+    } catch (cause) {
+      const message = cause instanceof Error ? cause.message : String(cause);
+      return `Error: ${message}`;
+    }
+  },
+};

packages/workflow-agent-builtin/src/tools/run-command.ts

@@ -0,0 +1,95 @@
+import { spawn } from "node:child_process";
+import { resolvePath } from "./path.js";
+import type { BuiltinTool } from "./types.js";
+
+const COMMAND_TIMEOUT_MS = 60_000;
+const MAX_OUTPUT_CHARS = 32_000;
+
+function isRecord(value: unknown): value is Record<string, unknown> {
+  return typeof value === "object" && value !== null && !Array.isArray(value);
+}
+
+function truncate(text: string, maxChars: number): string {
+  if (text.length <= maxChars) {
+    return text;
+  }
+  return `${text.slice(0, maxChars)}\n...(truncated)`;
+}
+
+function runShell(
+  command: string,
+  cwd: string,
+): Promise<{ stdout: string; stderr: string; code: number }> {
+  return new Promise((resolve, reject) => {
+    const child = spawn(command, {
+      cwd,
+      env: process.env,
+      shell: true,
+      stdio: ["ignore", "pipe", "pipe"],
+    });
+
+    let stdout = "";
+    let stderr = "";
+    child.stdout?.on("data", (chunk: Buffer) => {
+      stdout += chunk.toString();
+    });
+    child.stderr?.on("data", (chunk: Buffer) => {
+      stderr += chunk.toString();
+    });
+
+    const timer = setTimeout(() => {
+      child.kill("SIGTERM");
+    }, COMMAND_TIMEOUT_MS);
+
+    child.on("error", (cause) => {
+      clearTimeout(timer);
+      const message = cause instanceof Error ? cause.message : String(cause);
+      reject(new Error(message));
+    });
+
+    child.on("close", (code) => {
+      clearTimeout(timer);
+      resolve({ stdout, stderr, code: code ?? 1 });
+    });
+  });
+}
+
+export const runCommandTool: BuiltinTool = {
+  name: "run_command",
+  description: "Run a shell command. Output is truncated to 32KB.",
+  parameters: {
+    type: "object",
+    required: ["command"],
+    properties: {
+      command: { type: "string", description: "Shell command to execute." },
+      cwd: {
+        type: "string",
+        description: "Optional working directory relative to workspace root.",
+      },
+    },
+    additionalProperties: false,
+  },
+  execute: async (args, ctx) => {
+    if (!isRecord(args) || typeof args.command !== "string") {
+      return "Error: command must be a string";
+    }
+    let workDir = ctx.cwd;
+    if (args.cwd !== undefined && args.cwd !== null) {
+      if (typeof args.cwd !== "string") {
+        return "Error: cwd must be a string";
+      }
+      workDir = resolvePath(ctx.cwd, args.cwd);
+    }
+    try {
+      const { stdout, stderr, code } = await runShell(args.command, workDir);
+      const out = truncate(
+        `exit_code: ${code}\n--- stdout ---\n${stdout}\n--- stderr ---\n${stderr}`,
+        MAX_OUTPUT_CHARS,
+      );
+      return out;
+    } catch (cause) {
+      const message = cause instanceof Error ? cause.message : String(cause);
+      return `Error: ${message}`;
+    }
+  },
+};

packages/workflow-agent-builtin/src/tools/types.ts

@@ -0,0 +1,13 @@
+import type { JSONSchema } from "@uncaged/json-cas";
+
+export type ToolContext = {
+  cwd: string;
+  storageRoot: string;
+};
+
+export type BuiltinTool = {
+  name: string;
+  description: string;
+  parameters: JSONSchema;
+  execute: (args: unknown, ctx: ToolContext) => Promise<string>;
+};

packages/workflow-agent-builtin/src/tools/write-file.ts

@@ -0,0 +1,36 @@
+import { mkdir, writeFile } from "node:fs/promises";
+import { dirname } from "node:path";
+import { resolvePath } from "./path.js";
+import type { BuiltinTool } from "./types.js";
+
+function isRecord(value: unknown): value is Record<string, unknown> {
+  return typeof value === "object" && value !== null && !Array.isArray(value);
+}
+
+export const writeFileTool: BuiltinTool = {
+  name: "write_file",
+  description: "Write UTF-8 text to a file in the workspace (creates parent directories).",
+  parameters: {
+    type: "object",
+    required: ["path", "content"],
+    properties: {
+      path: { type: "string", description: "Relative or absolute path within the workspace." },
+      content: { type: "string", description: "File contents to write." },
+    },
+    additionalProperties: false,
+  },
+  execute: async (args, ctx) => {
+    if (!isRecord(args) || typeof args.path !== "string" || typeof args.content !== "string") {
+      return "Error: path and content must be strings";
+    }
+    const resolved = resolvePath(ctx.cwd, args.path);
+    try {
+      await mkdir(dirname(resolved), { recursive: true });
+      await writeFile(resolved, args.content, "utf8");
+      return `Wrote ${args.content.length} bytes to ${args.path}`;
+    } catch (cause) {
+      const message = cause instanceof Error ? cause.message : String(cause);
+      return `Error: ${message}`;
+    }
+  },
+};

packages/workflow-agent-builtin/src/types.ts

@@ -0,0 +1,49 @@
+import type { ChatMessage } from "./llm/index.js";
+
+export type BuiltinToolCallRecord = {
+  id: string;
+  name: string;
+  args: string;
+};
+
+export type BuiltinToolResultRecord = {
+  toolCallId: string;
+  name: string;
+  content: string;
+};
+
+export type BuiltinLoopTurn = {
+  assistantContent: string | null;
+  toolCalls: BuiltinToolCallRecord[] | null;
+  toolResults: BuiltinToolResultRecord[] | null;
+};
+
+export type BuiltinSessionState = {
+  sessionId: string;
+  model: string;
+  startedAtMs: number;
+  messages: ChatMessage[];
+  turns: BuiltinLoopTurn[];
+};
+
+export type BuiltinTurnRole = "assistant" | "tool";
+
+export type BuiltinToolCall = {
+  name: string;
+  args: string;
+};
+
+export type BuiltinTurnPayload = {
+  role: BuiltinTurnRole;
+  content: string;
+  toolCalls: BuiltinToolCall[] | null;
+  reasoning: string | null;
+};
+
+export type BuiltinDetailPayload = {
+  sessionId: string;
+  model: string;
+  duration: number;
+  turnCount: number;
+  turns: string[];
+};

packages/workflow-agent-builtin/tsconfig.json

@@ -0,0 +1,9 @@
+{
+  "extends": "../../tsconfig.json",
+  "compilerOptions": {
+    "rootDir": "src",
+    "outDir": "dist"
+  },
+  "include": ["src"],
+  "references": [{ "path": "../workflow-agent-kit" }, { "path": "../workflow-util" }]
+}

packages/workflow-agent-claude-code/README.md

@@ -0,0 +1,91 @@
+# @uncaged/workflow-agent-claude-code
+
+`uwf-claude-code` agent — spawns the Claude Code CLI and captures session detail.
+
+## Overview
+
+Layer 3 agent implementation. Spawns the `claude` CLI with a composed system prompt (role definition, task, prior steps, edge prompt). Parses stream or JSON stdout, caches session IDs for multi-turn continuation, and stores raw output plus structured detail in CAS.
+
+**Dependencies:** `@uncaged/json-cas`, `@uncaged/workflow-agent-kit`
+
+## Installation
+
+Included as the `uwf-claude-code` binary when you install `@uncaged/workflow-agent-claude-code`:
+
+```bash
+bun add -g @uncaged/workflow-agent-claude-code
+```
+
+Requires the `claude` CLI on `PATH`.
+
+## CLI Usage
+
+Invoked by `uwf thread step`:
+
+```bash
+uwf-claude-code <thread-id> <role>
+```
+
+Configure or override the agent:
+
+```bash
+uwf setup --agent claude-code
+uwf thread step <thread-id> --agent uwf-claude-code
+```
+
+Environment variables set by the engine:
+
+| Variable | Purpose |
+|----------|---------|
+| `UWF_EDGE_PROMPT` | Moderator edge instruction for this step |
+
+## API
+
+All exports come from `src/index.ts`.
+
+### Agent factory
+
+```typescript
+function createClaudeCodeAgent(): () => Promise<void>
+function buildClaudeCodePrompt(ctx: AgentContext): string
+```
+
+### Session detail
+
+```typescript
+function parseClaudeCodeStreamOutput(stdout: string): ClaudeCodeParsedResult | null
+function parseClaudeCodeJsonOutput(stdout: string): ClaudeCodeParsedResult | null
+function storeClaudeCodeDetail(
+  store: Store,
+  parsed: ClaudeCodeParsedResult,
+  sessionId: string,
+): Promise<string>
+function storeClaudeCodeRawOutput(store: Store, rawOutput: string): Promise<string>
+```
+
+## Usage (library)
+
+```typescript
+import { createClaudeCodeAgent, buildClaudeCodePrompt } from "@uncaged/workflow-agent-claude-code";
+
+const main = createClaudeCodeAgent();
+void main();
+```
+
+## Internal Structure
+
+```
+src/
+├── index.ts
+├── cli.ts              Binary entrypoint
+├── claude-code.ts      createClaudeCodeAgent, buildClaudeCodePrompt, spawn logic
+├── session-detail.ts   Parse stdout, store CAS detail nodes
+├── schemas.ts          Claude Code detail CAS schemas
+└── types.ts            ClaudeCodeParsedResult, message shapes
+```
+
+## Configuration
+
+Uses session caching from `@uncaged/workflow-agent-kit` (`getCachedSessionId` / `setCachedSessionId`). No separate config file — relies on the Claude Code CLI's own authentication.
+
+Maximum turns per invocation: 90 (constant in `claude-code.ts`).

packages/workflow-agent-claude-code/__tests__/claude-code.test.ts

@@ -0,0 +1,71 @@
+import { describe, expect, test } from "bun:test";
+import type { AgentContext } from "@uncaged/workflow-agent-kit";
+import type { ThreadId } from "@uncaged/workflow-protocol";
+import { buildClaudeCodePrompt } from "../src/claude-code.js";
+
+function makeCtx(overrides: Partial<AgentContext> = {}): AgentContext {
+  return {
+    threadId: "01JTEST0000000000000000000" as ThreadId,
+    edgePrompt: "Proceed with the assigned role.",
+    isFirstVisit: true,
+    workflow: {
+      roles: {
+        developer: {
+          description: "TDD implementation per test spec",
+          goal: "Write code",
+          capabilities: ["coding"],
+          procedure: "1. Read spec\n2. Write code",
+          output: "List files changed",
+          frontmatter: "",
+        },
+      },
+      conditions: {},
+      graph: {},
+    },
+    role: "developer",
+    start: { prompt: "Fix the bug", workflowHash: "abc123", threadId: "t1" },
+    steps: [],
+    store: {} as AgentContext["store"],
+    outputFormatInstruction: "Use YAML frontmatter",
+    ...overrides,
+  };
+}
+
+describe("buildClaudeCodePrompt", () => {
+  test("assembles outputFormatInstruction + role prompt + task prompt", () => {
+    const result = buildClaudeCodePrompt(makeCtx());
+    expect(result).toMatch(/^Use YAML frontmatter/);
+    expect(result).toContain("Write code");
+    expect(result).toContain("## Task\nFix the bug");
+  });
+
+  test("includes previous steps as history summary", () => {
+    const ctx = makeCtx({
+      steps: [
+        {
+          role: "planner",
+          output: '{"plan":"do X"}',
+          agent: "hermes",
+          detail: "detail-1",
+          edgePrompt: "Create a plan.",
+        },
+      ],
+    });
+    const result = buildClaudeCodePrompt(ctx);
+    expect(result).toContain("## Previous Steps");
+    expect(result).toContain("Step 1: planner");
+    expect(result).toContain("do X");
+  });
+
+  test("omits history section when steps array is empty", () => {
+    const result = buildClaudeCodePrompt(makeCtx({ steps: [] }));
+    expect(result).not.toContain("## Previous Steps");
+  });
+
+  test("works without outputFormatInstruction", () => {
+    const result = buildClaudeCodePrompt(makeCtx({ outputFormatInstruction: "" }));
+    expect(result).not.toMatch(/^\s*\n/);
+    expect(result).toContain("Write code");
+    expect(result).toContain("## Task");
+  });
+});

packages/workflow-agent-claude-code/__tests__/session-detail.test.ts

@@ -0,0 +1,314 @@
+import { describe, expect, test } from "bun:test";
+import { createMemoryStore, walk } from "@uncaged/json-cas";
+import {
+  parseClaudeCodeJsonOutput,
+  parseClaudeCodeStreamOutput,
+  storeClaudeCodeDetail,
+  storeClaudeCodeRawOutput,
+} from "../src/session-detail.js";
+import type { ClaudeCodeParsedResult } from "../src/types.js";
+
+describe("parseClaudeCodeJsonOutput", () => {
+  test("parses valid claude -p --output-format json output", () => {
+    const stdout = JSON.stringify({
+      type: "result",
+      subtype: "success",
+      result: "Done fixing bug",
+      session_id: "75e2167f-abc",
+      num_turns: 3,
+      total_cost_usd: 0.08,
+      duration_ms: 10276,
+      stop_reason: "end_turn",
+      usage: { input_tokens: 100, output_tokens: 50 },
+    });
+    const parsed = parseClaudeCodeJsonOutput(stdout);
+    expect(parsed).not.toBeNull();
+    expect(parsed!.type).toBe("result");
+    expect(parsed!.subtype).toBe("success");
+    expect(parsed!.result).toBe("Done fixing bug");
+    expect(parsed!.sessionId).toBe("75e2167f-abc");
+    expect(parsed!.numTurns).toBe(3);
+    expect(parsed!.totalCostUsd).toBe(0.08);
+    expect(parsed!.durationMs).toBe(10276);
+    expect(parsed!.stopReason).toBe("end_turn");
+    expect(parsed!.usage.inputTokens).toBe(100);
+    expect(parsed!.usage.outputTokens).toBe(50);
+    expect(parsed!.turns).toEqual([]);
+  });
+
+  test("returns null for non-JSON output", () => {
+    const parsed = parseClaudeCodeJsonOutput("Some random text\nwithout JSON");
+    expect(parsed).toBeNull();
+  });
+
+  test("returns null when session_id is missing", () => {
+    const stdout = JSON.stringify({ type: "result", result: "hi", subtype: "success" });
+    const parsed = parseClaudeCodeJsonOutput(stdout);
+    expect(parsed).toBeNull();
+  });
+});
+
+describe("parseClaudeCodeStreamOutput", () => {
+  test("parses stream-json output with turns", () => {
+    const lines = [
+      JSON.stringify({
+        type: "system",
+        subtype: "init",
+        session_id: "sess-123",
+        model: "claude-sonnet-4.5",
+        tools: ["Bash", "Read"],
+      }),
+      JSON.stringify({
+        type: "assistant",
+        message: {
+          role: "assistant",
+          content: [
+            { type: "text", text: "I'll list the files." },
+            { type: "tool_use", id: "tool_1", name: "Bash", input: { command: "ls" } },
+          ],
+        },
+        session_id: "sess-123",
+      }),
+      JSON.stringify({
+        type: "user",
+        message: {
+          role: "user",
+          content: [{ type: "tool_result", tool_use_id: "tool_1", content: "file1.ts\nfile2.ts" }],
+        },
+        session_id: "sess-123",
+      }),
+      JSON.stringify({
+        type: "assistant",
+        message: {
+          role: "assistant",
+          content: [{ type: "text", text: "There are 2 files." }],
+        },
+        session_id: "sess-123",
+      }),
+      JSON.stringify({
+        type: "result",
+        subtype: "success",
+        result: "There are 2 files.",
+        session_id: "sess-123",
+        num_turns: 2,
+        total_cost_usd: 0.05,
+        duration_ms: 5000,
+        stop_reason: "end_turn",
+        usage: {
+          input_tokens: 200,
+          output_tokens: 30,
+          cache_read_input_tokens: 100,
+          cache_creation_input_tokens: 0,
+        },
+      }),
+    ];
+    const stdout = lines.join("\n");
+    const parsed = parseClaudeCodeStreamOutput(stdout);
+
+    expect(parsed).not.toBeNull();
+    expect(parsed!.model).toBe("claude-sonnet-4.5");
+    expect(parsed!.sessionId).toBe("sess-123");
+    expect(parsed!.result).toBe("There are 2 files.");
+    expect(parsed!.stopReason).toBe("end_turn");
+    expect(parsed!.usage.inputTokens).toBe(200);
+    expect(parsed!.usage.outputTokens).toBe(30);
+    expect(parsed!.usage.cacheReadInputTokens).toBe(100);
+
+    // Turns: assistant(text+tool), tool_result, assistant(text)
+    expect(parsed!.turns).toHaveLength(3);
+    expect(parsed!.turns[0]!.role).toBe("assistant");
+    expect(parsed!.turns[0]!.content).toBe("I'll list the files.");
+    expect(parsed!.turns[0]!.toolCalls).toHaveLength(1);
+    expect(parsed!.turns[0]!.toolCalls![0]!.name).toBe("Bash");
+    expect(parsed!.turns[1]!.role).toBe("tool_result");
+    expect(parsed!.turns[1]!.content).toBe("file1.ts\nfile2.ts");
+    expect(parsed!.turns[2]!.role).toBe("assistant");
+    expect(parsed!.turns[2]!.content).toBe("There are 2 files.");
+    expect(parsed!.turns[2]!.toolCalls).toBeNull();
+  });
+
+  test("returns null when no result line", () => {
+    const stdout = JSON.stringify({ type: "system", model: "test" });
+    expect(parseClaudeCodeStreamOutput(stdout)).toBeNull();
+  });
+
+  test("skips invalid JSON lines gracefully", () => {
+    const lines = [
+      "not json",
+      JSON.stringify({
+        type: "result",
+        subtype: "success",
+        result: "ok",
+        session_id: "s1",
+        num_turns: 1,
+        total_cost_usd: 0.01,
+        duration_ms: 1000,
+        stop_reason: "end_turn",
+        usage: {},
+      }),
+    ];
+    const parsed = parseClaudeCodeStreamOutput(lines.join("\n"));
+    expect(parsed).not.toBeNull();
+    expect(parsed!.result).toBe("ok");
+    expect(parsed!.turns).toHaveLength(0);
+  });
+});
+
+describe("parseClaudeCodeStreamOutput — helper extraction", () => {
+  test("processSystemLine sets model from system message", () => {
+    const lines = [
+      JSON.stringify({ type: "system", model: "claude-opus-4" }),
+      JSON.stringify({
+        type: "result",
+        subtype: "success",
+        result: "ok",
+        session_id: "s1",
+        num_turns: 0,
+        total_cost_usd: 0,
+        duration_ms: 0,
+        stop_reason: "end_turn",
+      }),
+    ];
+    const parsed = parseClaudeCodeStreamOutput(lines.join("\n"));
+    expect(parsed).not.toBeNull();
+    expect(parsed!.model).toBe("claude-opus-4");
+  });
+
+  test("processAssistantLine skips empty content", () => {
+    const lines = [
+      JSON.stringify({ type: "assistant", message: { role: "assistant", content: [] } }),
+      JSON.stringify({
+        type: "result",
+        subtype: "success",
+        result: "ok",
+        session_id: "s1",
+        num_turns: 0,
+        total_cost_usd: 0,
+        duration_ms: 0,
+        stop_reason: "end_turn",
+      }),
+    ];
+    const parsed = parseClaudeCodeStreamOutput(lines.join("\n"));
+    expect(parsed).not.toBeNull();
+    expect(parsed!.turns).toHaveLength(0);
+  });
+
+  test("processUserLine skips when no tool_result items", () => {
+    const lines = [
+      JSON.stringify({
+        type: "user",
+        message: { role: "user", content: [{ type: "text", text: "hi" }] },
+      }),
+      JSON.stringify({
+        type: "result",
+        subtype: "success",
+        result: "ok",
+        session_id: "s1",
+        num_turns: 0,
+        total_cost_usd: 0,
+        duration_ms: 0,
+        stop_reason: "end_turn",
+      }),
+    ];
+    const parsed = parseClaudeCodeStreamOutput(lines.join("\n"));
+    expect(parsed).not.toBeNull();
+    expect(parsed!.turns).toHaveLength(0);
+  });
+
+  test("turn indices are sequential across mixed assistant and user lines", () => {
+    const lines = [
+      JSON.stringify({
+        type: "assistant",
+        message: { role: "assistant", content: [{ type: "text", text: "A" }] },
+      }),
+      JSON.stringify({
+        type: "user",
+        message: { role: "user", content: [{ type: "tool_result", content: "R" }] },
+      }),
+      JSON.stringify({
+        type: "assistant",
+        message: { role: "assistant", content: [{ type: "text", text: "B" }] },
+      }),
+      JSON.stringify({
+        type: "result",
+        subtype: "success",
+        result: "ok",
+        session_id: "s1",
+        num_turns: 3,
+        total_cost_usd: 0,
+        duration_ms: 0,
+        stop_reason: "end_turn",
+      }),
+    ];
+    const parsed = parseClaudeCodeStreamOutput(lines.join("\n"));
+    expect(parsed).not.toBeNull();
+    expect(parsed!.turns).toHaveLength(3);
+    expect(parsed!.turns.map((t) => t.index)).toEqual([0, 1, 2]);
+  });
+});
+
+describe("storeClaudeCodeDetail", () => {
+  const baseParsed: ClaudeCodeParsedResult = {
+    type: "result",
+    subtype: "success",
+    result: "The answer",
+    sessionId: "abc-123",
+    numTurns: 5,
+    totalCostUsd: 0.12,
+    durationMs: 15000,
+    model: "claude-sonnet-4.5",
+    stopReason: "end_turn",
+    usage: {
+      inputTokens: 100,
+      outputTokens: 50,
+      cacheReadInputTokens: 0,
+      cacheCreationInputTokens: 0,
+    },
+    turns: [
+      { index: 0, role: "assistant", content: "hello", toolCalls: null },
+      { index: 1, role: "tool_result", content: "world", toolCalls: null },
+    ],
+  };
+
+  test("stores detail with per-turn CAS nodes", async () => {
+    const store = createMemoryStore();
+    const { detailHash, output, sessionId } = await storeClaudeCodeDetail(store, baseParsed);
+
+    expect(detailHash).toHaveLength(13);
+    expect(output).toBe("The answer");
+    expect(sessionId).toBe("abc-123");
+
+    const node = await store.get(detailHash);
+    expect(node).not.toBeNull();
+    expect(node!.payload.model).toBe("claude-sonnet-4.5");
+    expect(node!.payload.stopReason).toBe("end_turn");
+    expect(node!.payload.usage.inputTokens).toBe(100);
+    expect(node!.payload.turns).toHaveLength(2);
+
+    // Verify turn CAS nodes
+    const turn0 = await store.get(node!.payload.turns[0]);
+    expect(turn0).not.toBeNull();
+    expect(turn0!.payload.role).toBe("assistant");
+    expect(turn0!.payload.content).toBe("hello");
+  });
+
+  test("detail node is walkable from root", async () => {
+    const store = createMemoryStore();
+    const { detailHash } = await storeClaudeCodeDetail(store, baseParsed);
+    const visited: string[] = [];
+    walk(store, detailHash, (hash) => visited.push(hash));
+    expect(visited.length).toBeGreaterThan(0);
+  });
+});
+
+describe("storeClaudeCodeRawOutput", () => {
+  test("stores raw text when JSON parsing fails", async () => {
+    const store = createMemoryStore();
+    const rawText = "Claude produced plain text without JSON";
+    const hash = await storeClaudeCodeRawOutput(store, rawText);
+    expect(hash).toHaveLength(13);
+    const node = await store.get(hash);
+    expect(node).not.toBeNull();
+    expect(node!.payload.text).toBe(rawText);
+  });
+});

packages/workflow-agent-claude-code/package.json

@@ -0,0 +1,33 @@
+{
+  "name": "@uncaged/workflow-agent-claude-code",
+  "version": "0.1.0",
+  "files": [
+    "src",
+    "dist",
+    "package.json"
+  ],
+  "type": "module",
+  "bin": {
+    "uwf-claude-code": "./src/cli.ts"
+  },
+  "exports": {
+    ".": {
+      "bun": "./src/index.ts",
+      "types": "./dist/index.d.ts",
+      "import": "./dist/index.js"
+    }
+  },
+  "scripts": {
+    "test": "bun test"
+  },
+  "dependencies": {
+    "@uncaged/json-cas": "^0.4.0",
+    "@uncaged/workflow-agent-kit": "workspace:^"
+  },
+  "devDependencies": {
+    "typescript": "^5.8.3"
+  },
+  "publishConfig": {
+    "access": "public"
+  }
+}

packages/workflow-agent-claude-code/src/claude-code.ts

@@ -0,0 +1,193 @@
+import { spawn } from "node:child_process";
+import type { Store } from "@uncaged/json-cas";
+import {
+  type AgentContext,
+  type AgentRunResult,
+  buildRolePrompt,
+  createAgent,
+  getCachedSessionId,
+  setCachedSessionId,
+} from "@uncaged/workflow-agent-kit";
+import { createLogger } from "@uncaged/workflow-util";
+
+import { parseClaudeCodeStreamOutput, storeClaudeCodeDetail } from "./session-detail.js";
+
+const log = createLogger({ sink: { kind: "stderr" } });
+
+const CLAUDE_COMMAND = "claude";
+const CLAUDE_MAX_TURNS = 90;
+const CLAUDE_MODEL = process.env.CLAUDE_MODEL ?? null;
+
+function buildHistorySummary(steps: AgentContext["steps"]): string {
+  if (steps.length === 0) {
+    return "";
+  }
+
+  const lines: string[] = ["## Previous Steps"];
+  for (let i = 0; i < steps.length; i++) {
+    const step = steps[i];
+    if (step === undefined) {
+      continue;
+    }
+    lines.push("");
+    lines.push(`### Step ${i + 1}: ${step.role}`);
+    lines.push(`Output: ${JSON.stringify(step.output)}`);
+    lines.push(`Agent: ${step.agent}`);
+  }
+  return lines.join("\n");
+}
+
+/** Assemble system prompt, task, and prior step outputs for Claude Code. */
+export function buildClaudeCodePrompt(ctx: AgentContext): string {
+  const roleDef = ctx.workflow.roles[ctx.role];
+  const rolePrompt = roleDef !== undefined ? buildRolePrompt(roleDef) : "";
+  const parts: string[] = [];
+  if (ctx.outputFormatInstruction !== undefined && ctx.outputFormatInstruction !== "") {
+    parts.push(ctx.outputFormatInstruction, "");
+  }
+  parts.push(rolePrompt, "", "## Task", ctx.start.prompt);
+  const historyBlock = buildHistorySummary(ctx.steps);
+  if (historyBlock !== "") {
+    parts.push("", historyBlock);
+  }
+  parts.push("", "## Current Instruction", "", ctx.edgePrompt);
+  return parts.join("\n");
+}
+
+function spawnClaude(args: string[]): Promise<{ stdout: string; stderr: string }> {
+  return new Promise((resolve, reject) => {
+    const child = spawn(CLAUDE_COMMAND, args, {
+      env: process.env,
+      shell: false,
+      stdio: ["ignore", "pipe", "pipe"],
+    });
+
+    let stdout = "";
+    let stderr = "";
+    child.stdout?.on("data", (chunk: Buffer) => {
+      stdout += chunk.toString();
+    });
+    child.stderr?.on("data", (chunk: Buffer) => {
+      stderr += chunk.toString();
+    });
+
+    child.on("error", (cause) => {
+      const message = cause instanceof Error ? cause.message : String(cause);
+      reject(new Error(`claude spawn failed: ${message}`));
+    });
+
+    child.on("close", (code) => {
+      if (code === 0) {
+        resolve({ stdout, stderr });
+        return;
+      }
+      const detail = stderr.trim() !== "" ? ` stderr=${stderr.trim()}` : "";
+      reject(new Error(`claude exited with code ${code ?? "null"}${detail}`));
+    });
+  });
+}
+
+function spawnClaudeRun(prompt: string): Promise<{ stdout: string; stderr: string }> {
+  const args = [
+    "-p",
+    prompt,
+    "--output-format",
+    "stream-json",
+    "--verbose",
+    "--dangerously-skip-permissions",
+    "--max-turns",
+    String(CLAUDE_MAX_TURNS),
+  ];
+  if (CLAUDE_MODEL !== null) {
+    args.push("--model", CLAUDE_MODEL);
+  }
+  return spawnClaude(args);
+}
+
+function spawnClaudeResume(
+  sessionId: string,
+  message: string,
+): Promise<{ stdout: string; stderr: string }> {
+  const args = [
+    "-p",
+    message,
+    "--resume",
+    sessionId,
+    "--output-format",
+    "stream-json",
+    "--verbose",
+    "--dangerously-skip-permissions",
+    "--max-turns",
+    String(CLAUDE_MAX_TURNS),
+  ];
+  if (CLAUDE_MODEL !== null) {
+    args.push("--model", CLAUDE_MODEL);
+  }
+  return spawnClaude(args);
+}
+
+async function processClaudeOutput(stdout: string, store: Store): Promise<AgentRunResult> {
+  const parsed = parseClaudeCodeStreamOutput(stdout);
+
+  if (parsed !== null) {
+    const { detailHash, output, sessionId } = await storeClaudeCodeDetail(store, parsed);
+    return { output, detailHash, sessionId };
+  }
+
+  throw new Error(
+    `Claude Code returned unparseable output (first 200 chars): ${stdout.slice(0, 200)}`,
+  );
+}
+
+async function runClaudeCode(ctx: AgentContext): Promise<AgentRunResult> {
+  const fullPrompt = buildClaudeCodePrompt(ctx);
+
+  log("K7R2M4N8", `prompt for role=${ctx.role} (length=${fullPrompt.length}):\n${fullPrompt}`);
+
+  // Try resuming a cached session for re-entry scenarios (e.g. reviewer reject → developer re-entry).
+  if (!ctx.isFirstVisit) {
+    const cachedSessionId = await getCachedSessionId("claude-code", ctx.threadId, ctx.role);
+    if (cachedSessionId !== null) {
+      try {
+        const { stdout } = await spawnClaudeResume(cachedSessionId, fullPrompt);
+        const result = await processClaudeOutput(stdout, ctx.store);
+        if (result.sessionId !== undefined && result.sessionId !== "") {
+          await setCachedSessionId("claude-code", ctx.threadId, ctx.role, result.sessionId);
+        }
+        return result;
+      } catch (err) {
+        log(
+          "5VKR8N3Q",
+          "resume failed for session %s, falling back to fresh run: %s",
+          cachedSessionId,
+          err,
+        );
+      }
+    }
+  }
+
+  const { stdout } = await spawnClaudeRun(fullPrompt);
+  const result = await processClaudeOutput(stdout, ctx.store);
+  if (result.sessionId !== undefined && result.sessionId !== "") {
+    await setCachedSessionId("claude-code", ctx.threadId, ctx.role, result.sessionId);
+  }
+  return result;
+}
+
+async function continueClaudeCode(
+  sessionId: string,
+  message: string,
+  store: Store,
+): Promise<AgentRunResult> {
+  const { stdout } = await spawnClaudeResume(sessionId, message);
+  return processClaudeOutput(stdout, store);
+}
+
+/** Agent CLI factory: parses argv, runs Claude Code, extracts output, writes StepNode. */
+export function createClaudeCodeAgent(): () => Promise<void> {
+  return createAgent({
+    name: "claude-code",
+    run: runClaudeCode,
+    continue: continueClaudeCode,
+  });
+}

packages/workflow-agent-claude-code/src/cli.ts

@@ -0,0 +1,6 @@
+#!/usr/bin/env bun
+
+import { createClaudeCodeAgent } from "./claude-code.js";
+
+const main = createClaudeCodeAgent();
+void main();

packages/workflow-agent-claude-code/src/index.ts

@@ -0,0 +1,7 @@
+export { buildClaudeCodePrompt, createClaudeCodeAgent } from "./claude-code.js";
+export {
+  parseClaudeCodeJsonOutput,
+  parseClaudeCodeStreamOutput,
+  storeClaudeCodeDetail,
+  storeClaudeCodeRawOutput,
+} from "./session-detail.js";

packages/workflow-agent-claude-code/src/schemas.ts

@@ -0,0 +1,64 @@
+import type { JSONSchema } from "@uncaged/json-cas";
+
+export const CLAUDE_CODE_DETAIL_SCHEMA: JSONSchema = {
+  title: "claude-code-detail",
+  type: "object",
+  required: [
+    "sessionId",
+    "model",
+    "subtype",
+    "durationMs",
+    "numTurns",
+    "totalCostUsd",
+    "stopReason",
+    "usage",
+    "turns",
+  ],
+  properties: {
+    sessionId: { type: "string" },
+    model: { type: "string" },
+    subtype: { type: "string" },
+    durationMs: { type: "integer" },
+    numTurns: { type: "integer" },
+    totalCostUsd: { type: "number" },
+    stopReason: { type: "string" },
+    usage: {
+      type: "object",
+      properties: {
+        inputTokens: { type: "integer" },
+        outputTokens: { type: "integer" },
+        cacheReadInputTokens: { type: "integer" },
+        cacheCreationInputTokens: { type: "integer" },
+      },
+      required: ["inputTokens", "outputTokens", "cacheReadInputTokens", "cacheCreationInputTokens"],
+    },
+    turns: {
+      type: "array",
+      items: { type: "string", format: "cas_ref" },
+    },
+  },
+  additionalProperties: false,
+};
+
+export const CLAUDE_CODE_TURN_SCHEMA: JSONSchema = {
+  title: "claude-code-turn",
+  type: "object",
+  required: ["index", "role", "content", "toolCalls"],
+  properties: {
+    index: { type: "integer" },
+    role: { type: "string" },
+    content: { type: "string" },
+    toolCalls: {},
+  },
+  additionalProperties: false,
+};
+
+export const CLAUDE_CODE_RAW_OUTPUT_SCHEMA: JSONSchema = {
+  title: "claude-code-raw-output",
+  type: "object",
+  required: ["text"],
+  properties: {
+    text: { type: "string" },
+  },
+  additionalProperties: false,
+};

packages/workflow-agent-claude-code/src/session-detail.ts

@@ -0,0 +1,263 @@
+import { bootstrap, putSchema, type Store } from "@uncaged/json-cas";
+
+import {
+  CLAUDE_CODE_DETAIL_SCHEMA,
+  CLAUDE_CODE_RAW_OUTPUT_SCHEMA,
+  CLAUDE_CODE_TURN_SCHEMA,
+} from "./schemas.js";
+import type {
+  ClaudeCodeDetailPayload,
+  ClaudeCodeParsedResult,
+  ClaudeCodeToolCall,
+  ClaudeCodeTurnPayload,
+} from "./types.js";
+
+function isRecord(value: unknown): value is Record<string, unknown> {
+  return typeof value === "object" && value !== null && !Array.isArray(value);
+}
+
+function safeNumber(v: unknown, fallback = 0): number {
+  return typeof v === "number" ? v : fallback;
+}
+
+function safeString(v: unknown, fallback = ""): string {
+  return typeof v === "string" ? v : fallback;
+}
+
+/**
+ * Extract tool calls from an assistant message content array.
+ */
+function extractToolCalls(content: unknown[]): ClaudeCodeToolCall[] {
+  const calls: ClaudeCodeToolCall[] = [];
+  for (const item of content) {
+    if (isRecord(item) && item.type === "tool_use" && typeof item.name === "string") {
+      calls.push({
+        name: item.name,
+        input: typeof item.input === "string" ? item.input : JSON.stringify(item.input ?? {}),
+      });
+    }
+  }
+  return calls;
+}
+
+/**
+ * Extract text content from a message content array.
+ */
+function extractTextContent(content: unknown[]): string {
+  const texts: string[] = [];
+  for (const item of content) {
+    if (isRecord(item) && item.type === "text" && typeof item.text === "string") {
+      texts.push(item.text);
+    }
+  }
+  return texts.join("\n");
+}
+
+/**
+ * Extract tool result content from a user message content array.
+ */
+function extractToolResultContent(content: unknown[]): string {
+  const results: string[] = [];
+  for (const item of content) {
+    if (isRecord(item) && item.type === "tool_result") {
+      const text = typeof item.content === "string" ? item.content : "";
+      results.push(text);
+    }
+  }
+  return results.join("\n");
+}
+
+type ParseState = {
+  turns: ClaudeCodeTurnPayload[];
+  resultLine: Record<string, unknown> | null;
+  model: string;
+  turnIndex: number;
+};
+
+function processSystemLine(parsed: Record<string, unknown>, state: ParseState): void {
+  if (typeof parsed.model === "string") {
+    state.model = parsed.model;
+  }
+}
+
+function processAssistantLine(parsed: Record<string, unknown>, state: ParseState): void {
+  if (!isRecord(parsed.message)) return;
+  const content = Array.isArray(parsed.message.content) ? parsed.message.content : [];
+  const textContent = extractTextContent(content as unknown[]);
+  const toolCalls = extractToolCalls(content as unknown[]);
+  if (textContent !== "" || toolCalls.length > 0) {
+    state.turns.push({
+      index: state.turnIndex++,
+      role: "assistant",
+      content: textContent,
+      toolCalls: toolCalls.length > 0 ? toolCalls : null,
+    });
+  }
+}
+
+function processUserLine(parsed: Record<string, unknown>, state: ParseState): void {
+  if (!isRecord(parsed.message)) return;
+  const content = Array.isArray(parsed.message.content) ? parsed.message.content : [];
+  const resultContent = extractToolResultContent(content as unknown[]);
+  if (resultContent !== "") {
+    state.turns.push({
+      index: state.turnIndex++,
+      role: "tool_result",
+      content: resultContent,
+      toolCalls: null,
+    });
+  }
+}
+
+function processLine(line: string, state: ParseState): void {
+  let parsed: unknown;
+  try {
+    parsed = JSON.parse(line);
+  } catch {
+    return;
+  }
+  if (!isRecord(parsed)) return;
+  const type = parsed.type;
+  if (type === "system") processSystemLine(parsed, state);
+  else if (type === "assistant") processAssistantLine(parsed, state);
+  else if (type === "user") processUserLine(parsed, state);
+  else if (type === "result") state.resultLine = parsed;
+}
+
+function assembleResult(state: ParseState): ClaudeCodeParsedResult | null {
+  if (state.resultLine === null) return null;
+  const sessionId = state.resultLine.session_id;
+  const result = state.resultLine.result;
+  const subtype = state.resultLine.subtype;
+  if (typeof sessionId !== "string" || typeof result !== "string" || typeof subtype !== "string") {
+    return null;
+  }
+  const usage = isRecord(state.resultLine.usage) ? state.resultLine.usage : {};
+  return {
+    type: safeString(state.resultLine.type, "result"),
+    subtype: subtype as ClaudeCodeParsedResult["subtype"],
+    result,
+    sessionId,
+    numTurns: safeNumber(state.resultLine.num_turns),
+    totalCostUsd: safeNumber(state.resultLine.total_cost_usd),
+    durationMs: safeNumber(state.resultLine.duration_ms),
+    model: state.model,
+    stopReason: safeString(state.resultLine.stop_reason),
+    usage: {
+      inputTokens: safeNumber(usage.input_tokens),
+      outputTokens: safeNumber(usage.output_tokens),
+      cacheReadInputTokens: safeNumber(usage.cache_read_input_tokens),
+      cacheCreationInputTokens: safeNumber(usage.cache_creation_input_tokens),
+    },
+    turns: state.turns,
+  };
+}
+
+/**
+ * Parse Claude Code stream-json (NDJSON) output.
+ * Each line is a JSON object with type: "system" | "assistant" | "user" | "result".
+ */
+export function parseClaudeCodeStreamOutput(stdout: string): ClaudeCodeParsedResult | null {
+  const lines = stdout.trim().split("\n");
+  const state: ParseState = { turns: [], resultLine: null, model: "", turnIndex: 0 };
+  for (const line of lines) {
+    processLine(line, state);
+  }
+  return assembleResult(state);
+}
+
+/**
+ * Legacy: parse Claude Code plain JSON output (non-streaming).
+ * Falls back when stream-json is not available.
+ */
+export function parseClaudeCodeJsonOutput(stdout: string): ClaudeCodeParsedResult | null {
+  let parsed: unknown;
+  try {
+    parsed = JSON.parse(stdout.trim());
+  } catch {
+    return null;
+  }
+
+  if (!isRecord(parsed)) return null;
+
+  const sessionId = parsed.session_id;
+  const result = parsed.result;
+  const subtype = parsed.subtype;
+
+  if (typeof sessionId !== "string" || typeof result !== "string" || typeof subtype !== "string") {
+    return null;
+  }
+
+  const usage = isRecord(parsed.usage) ? parsed.usage : {};
+
+  return {
+    type: safeString(parsed.type, "result"),
+    subtype: subtype as ClaudeCodeParsedResult["subtype"],
+    result,
+    sessionId,
+    numTurns: safeNumber(parsed.num_turns),
+    totalCostUsd: safeNumber(parsed.total_cost_usd),
+    durationMs: safeNumber(parsed.duration_ms),
+    model: "",
+    stopReason: safeString(parsed.stop_reason),
+    usage: {
+      inputTokens: safeNumber(usage.input_tokens),
+      outputTokens: safeNumber(usage.output_tokens),
+      cacheReadInputTokens: safeNumber(usage.cache_read_input_tokens),
+      cacheCreationInputTokens: safeNumber(usage.cache_creation_input_tokens),
+    },
+    turns: [],
+  };
+}
+
+type ClaudeCodeSchemaHashes = {
+  detail: string;
+  turn: string;
+  rawOutput: string;
+};
+
+async function registerSchemas(store: Store): Promise<ClaudeCodeSchemaHashes> {
+  await bootstrap(store);
+  const [detail, turn, rawOutput] = await Promise.all([
+    putSchema(store, CLAUDE_CODE_DETAIL_SCHEMA),
+    putSchema(store, CLAUDE_CODE_TURN_SCHEMA),
+    putSchema(store, CLAUDE_CODE_RAW_OUTPUT_SCHEMA),
+  ]);
+  return { detail, turn, rawOutput };
+}
+
+/** Store parsed Claude Code result with per-turn breakdown as CAS detail nodes. */
+export async function storeClaudeCodeDetail(
+  store: Store,
+  parsed: ClaudeCodeParsedResult,
+): Promise<{ detailHash: string; output: string; sessionId: string }> {
+  const schemas = await registerSchemas(store);
+
+  // Store each turn as an individual CAS node
+  const turnHashes: string[] = [];
+  for (const turn of parsed.turns) {
+    const hash = await store.put(schemas.turn, turn);
+    turnHashes.push(hash);
+  }
+
+  const detail: ClaudeCodeDetailPayload = {
+    sessionId: parsed.sessionId,
+    model: parsed.model,
+    subtype: parsed.subtype,
+    durationMs: parsed.durationMs,
+    numTurns: parsed.numTurns,
+    totalCostUsd: parsed.totalCostUsd,
+    stopReason: parsed.stopReason,
+    usage: parsed.usage,
+    turns: turnHashes,
+  };
+
+  const detailHash = await store.put(schemas.detail, detail);
+  return { detailHash, output: parsed.result, sessionId: parsed.sessionId };
+}
+
+/** Fallback: store raw text output when JSON parsing fails. */
+export async function storeClaudeCodeRawOutput(store: Store, rawOutput: string): Promise<string> {
+  const schemas = await registerSchemas(store);
+  return store.put(schemas.rawOutput, { text: rawOutput });
+}

packages/workflow-agent-claude-code/src/types.ts

@@ -0,0 +1,53 @@
+export type ClaudeCodeResultSubtype = "success" | "error_max_turns" | "error_budget";
+
+/** A single tool call within an assistant turn. */
+export type ClaudeCodeToolCall = {
+  name: string;
+  input: string;
+};
+
+/** A single turn (assistant text, tool use, or tool result). */
+export type ClaudeCodeTurnPayload = {
+  index: number;
+  role: "assistant" | "tool_result";
+  content: string;
+  toolCalls: ClaudeCodeToolCall[] | null;
+};
+
+/** Top-level detail stored as CAS node. */
+export type ClaudeCodeDetailPayload = {
+  sessionId: string;
+  model: string;
+  subtype: string;
+  durationMs: number;
+  numTurns: number;
+  totalCostUsd: number;
+  stopReason: string;
+  usage: {
+    inputTokens: number;
+    outputTokens: number;
+    cacheReadInputTokens: number;
+    cacheCreationInputTokens: number;
+  };
+  turns: string[]; // CAS hashes of ClaudeCodeTurnPayload
+};
+
+/** Intermediate parsed result from stream-json output. */
+export type ClaudeCodeParsedResult = {
+  type: string;
+  subtype: ClaudeCodeResultSubtype;
+  result: string;
+  sessionId: string;
+  numTurns: number;
+  totalCostUsd: number;
+  durationMs: number;
+  model: string;
+  stopReason: string;
+  usage: {
+    inputTokens: number;
+    outputTokens: number;
+    cacheReadInputTokens: number;
+    cacheCreationInputTokens: number;
+  };
+  turns: ClaudeCodeTurnPayload[];
+};

packages/workflow-agent-claude-code/tsconfig.json

@@ -0,0 +1,6 @@
+{
+  "extends": "../../tsconfig.json",
+  "compilerOptions": { "rootDir": "src", "outDir": "dist" },
+  "include": ["src"],
+  "references": [{ "path": "../workflow-agent-kit" }]
+}

packages/workflow-agent-hermes/README.md

@@ -0,0 +1,90 @@
+# @uncaged/workflow-agent-hermes
+
+`uwf-hermes` agent — spawns Hermes chat via ACP and captures session detail.
+
+## Overview
+
+Layer 3 agent implementation. Wraps the Hermes CLI using the Agent Client Protocol (ACP). On first visit to a role it sends a composed prompt (role definition, task, history, edge prompt); on continuation it resumes the cached session. Session transcripts and raw output are stored as CAS detail nodes.
+
+**Dependencies:** `@uncaged/json-cas`, `@uncaged/workflow-agent-kit`, `@uncaged/workflow-protocol`, `@uncaged/workflow-util`
+
+## Installation
+
+Included as the `uwf-hermes` binary when you install `@uncaged/workflow-agent-hermes`:
+
+```bash
+bun add -g @uncaged/workflow-agent-hermes
+```
+
+Requires the `hermes` CLI on `PATH`.
+
+## CLI Usage
+
+Invoked by `uwf thread step` (not typically run directly):
+
+```bash
+uwf-hermes <thread-id> <role>
+```
+
+Environment variables set by the engine:
+
+| Variable | Purpose |
+|----------|---------|
+| `UWF_EDGE_PROMPT` | Moderator edge instruction for this step |
+
+Configure as the default agent via `uwf setup --agent hermes`.
+
+Override per step:
+
+```bash
+uwf thread step <thread-id> --agent uwf-hermes
+```
+
+## API
+
+All exports come from `src/index.ts`.
+
+### Agent factory
+
+```typescript
+function createHermesAgent(): () => Promise<void>
+function buildHermesPrompt(ctx: AgentContext): string
+```
+
+### ACP client
+
+```typescript
+class HermesAcpClient {
+  // Spawns hermes, handles JSON-RPC over stdio
+}
+```
+
+## Usage (library)
+
+```typescript
+import { createHermesAgent, buildHermesPrompt } from "@uncaged/workflow-agent-hermes";
+
+// CLI entry (src/cli.ts):
+const main = createHermesAgent();
+void main();
+```
+
+## Internal Structure
+
+```
+src/
+├── index.ts
+├── cli.ts              Binary entrypoint
+├── hermes.ts           createHermesAgent, buildHermesPrompt
+├── acp-client.ts       HermesAcpClient — ACP JSON-RPC over stdio
+├── session-cache.ts    Session ID cache (re-exports kit helpers + isResumeDisabled)
+├── session-detail.ts   Parse Hermes session JSON, store CAS detail nodes
+├── schemas.ts          Hermes detail CAS schemas
+└── types.ts            HermesSessionJson, HermesSessionMessage
+```
+
+## Configuration
+
+Uses workflow config from `~/.uncaged/workflow/config.yaml` (via agent-kit). Hermes session files are stored under the workflow storage root (see `session-detail.ts`).
+
+Set `UWF_HERMES_NO_RESUME=1` to disable session resume (see `isResumeDisabled` in `session-cache.ts`).

packages/workflow-agent-hermes/__tests__/acp-client.test.ts

@@ -0,0 +1,168 @@
+import { afterEach, beforeEach, describe, expect, it } from "bun:test";
+
+import { HermesAcpClient } from "../src/acp-client.js";
+
+const UUID_RE = /^[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}$/i;
+
+describe("handleSessionUpdate — helper extraction", () => {
+  let client: HermesAcpClient;
+
+  beforeEach(() => {
+    client = new HermesAcpClient();
+  });
+
+  afterEach(async () => {
+    await client.close();
+  });
+
+  it("agent_message_chunk accumulates text in messageChunks", () => {
+    (client as any).handleSessionUpdate({
+      sessionUpdate: "agent_message_chunk",
+      content: { type: "text", text: "hello" },
+    });
+    (client as any).handleSessionUpdate({
+      sessionUpdate: "agent_message_chunk",
+      content: { type: "text", text: " world" },
+    });
+    expect((client as any).messageChunks).toEqual(["hello", " world"]);
+  });
+
+  it("agent_thought_chunk accumulates reasoning in reasoningChunks", () => {
+    (client as any).handleSessionUpdate({
+      sessionUpdate: "agent_thought_chunk",
+      content: { type: "text", text: "thinking" },
+    });
+    expect((client as any).reasoningChunks).toEqual(["thinking"]);
+  });
+
+  it("tool_call registers a pending tool and flushes message chunks", () => {
+    (client as any).messageChunks = ["pre-tool text"];
+    (client as any).handleSessionUpdate({
+      sessionUpdate: "tool_call",
+      title: "Bash",
+      rawInput: { command: "ls" },
+      toolCallId: "tc-1",
+    });
+    expect((client as any).pendingTools.get("tc-1")).toEqual({
+      name: "Bash",
+      args: JSON.stringify({ command: "ls" }),
+    });
+    expect((client as any).messageChunks).toEqual([]);
+    expect((client as any).messages).toHaveLength(1);
+    expect((client as any).messages[0].role).toBe("assistant");
+  });
+
+  it("tool_call_update completed pushes tool_call and tool messages", () => {
+    (client as any).pendingTools.set("tc-2", { name: "Read", args: '{"path":"/foo"}' });
+    (client as any).handleSessionUpdate({
+      sessionUpdate: "tool_call_update",
+      status: "completed",
+      toolCallId: "tc-2",
+      rawOutput: "file contents",
+    });
+    const msgs = (client as any).messages as Array<{
+      role: string;
+      tool_calls: unknown;
+      content: string | null;
+    }>;
+    expect(msgs).toHaveLength(2);
+    expect(msgs[0].role).toBe("assistant");
+    expect(msgs[0].tool_calls).toEqual([
+      { function: { name: "Read", arguments: '{"path":"/foo"}' } },
+    ]);
+    expect(msgs[1].role).toBe("tool");
+    expect(msgs[1].content).toBe("file contents");
+    expect((client as any).pendingTools.has("tc-2")).toBe(false);
+  });
+
+  it("tool_call_update with non-string rawOutput JSON-stringifies it", () => {
+    (client as any).pendingTools.set("tc-3", { name: "Fetch", args: "" });
+    (client as any).handleSessionUpdate({
+      sessionUpdate: "tool_call_update",
+      status: "completed",
+      toolCallId: "tc-3",
+      rawOutput: { html: "<p>page</p>" },
+    });
+    const msgs = (client as any).messages as Array<{ role: string; content: string | null }>;
+    expect(msgs[1].content).toBe(JSON.stringify({ html: "<p>page</p>" }));
+  });
+
+  it("unknown updateType is a no-op", () => {
+    (client as any).handleSessionUpdate({ sessionUpdate: "unknown_type", data: {} });
+    expect((client as any).messages).toHaveLength(0);
+    expect((client as any).messageChunks).toHaveLength(0);
+  });
+});
+
+describe("HermesAcpClient", () => {
+  let client: HermesAcpClient;
+
+  beforeEach(() => {
+    client = new HermesAcpClient();
+  });
+
+  afterEach(async () => {
+    await client.close();
+  });
+
+  it(
+    "connect() returns a UUID sessionId",
+    async () => {
+      const sessionId = await client.connect(process.cwd());
+      expect(typeof sessionId).toBe("string");
+      expect(sessionId).toMatch(UUID_RE);
+    },
+    { timeout: 2 * 60 * 1000 },
+  );
+
+  it(
+    "prompt() returns a non-empty text response",
+    async () => {
+      await client.connect(process.cwd());
+      const result = await client.prompt("Reply with exactly the word: PONG");
+      expect(typeof result.text).toBe("string");
+      expect(result.text.length).toBeGreaterThan(0);
+      expect(typeof result.sessionId).toBe("string");
+      expect(result.sessionId).toMatch(UUID_RE);
+    },
+    { timeout: 2 * 60 * 1000 },
+  );
+
+  it(
+    "prompt() can be called twice on the same session (resume)",
+    async () => {
+      await client.connect(process.cwd());
+
+      const first = await client.prompt("Say the word ALPHA and nothing else.");
+      expect(first.text.length).toBeGreaterThan(0);
+
+      const second = await client.prompt("Now say the word BETA and nothing else.");
+      expect(second.text.length).toBeGreaterThan(0);
+
+      expect(first.sessionId).toBe(second.sessionId);
+    },
+    { timeout: 2 * 60 * 1000 },
+  );
+
+  // TODO(#435): flaky — depends on live LLM; mock or move to integration suite
+  it.skip(
+    "prompt() collects structured messages including tool calls",
+    async () => {
+      await client.connect(process.cwd());
+      const result = await client.prompt("Run this command: echo TOOL_DETAIL_TEST");
+      expect(result.messages.length).toBeGreaterThan(0);
+      // Should have at least one tool message (the echo command)
+      const toolMessages = result.messages.filter((m) => m.role === "tool");
+      expect(toolMessages.length).toBeGreaterThan(0);
+      // Tool message should contain the output
+      const toolContent = toolMessages[0]?.content ?? "";
+      expect(toolContent).toContain("TOOL_DETAIL_TEST");
+      // Should have assistant messages with tool_calls
+      const assistantWithTools = result.messages.filter(
+        (m) => m.role === "assistant" && m.tool_calls !== null,
+      );
+      expect(assistantWithTools.length).toBeGreaterThan(0);
+    },
+    { timeout: 2 * 60 * 1000 },
+  );
+});

packages/workflow-agent-hermes/__tests__/hermes-prompt.test.ts

@@ -0,0 +1,98 @@
+import { describe, expect, test } from "bun:test";
+import type { AgentContext } from "@uncaged/workflow-agent-kit";
+import type { ThreadId } from "@uncaged/workflow-protocol";
+import { buildHermesPrompt } from "../src/hermes.js";
+
+function makeCtx(overrides: Partial<AgentContext> = {}): AgentContext {
+  return {
+    threadId: "01JTEST0000000000000000000" as ThreadId,
+    edgePrompt: "Proceed with the assigned role.",
+    isFirstVisit: true,
+    workflow: {
+      roles: {
+        developer: {
+          description: "TDD implementation per test spec",
+          goal: "Write code",
+          capabilities: ["coding"],
+          procedure: "1. Read spec\n2. Write code",
+          output: "List files changed",
+          frontmatter: "",
+        },
+      },
+      conditions: {},
+      graph: {},
+    },
+    role: "developer",
+    start: { prompt: "Fix the bug", workflowHash: "abc123", threadId: "t1" },
+    steps: [],
+    store: {} as AgentContext["store"],
+    outputFormatInstruction: "Use YAML frontmatter",
+    ...overrides,
+  };
+}
+
+describe("buildHermesPrompt", () => {
+  test("first visit uses full role prompt and includes moderator instruction", () => {
+    const result = buildHermesPrompt(
+      makeCtx({ edgePrompt: "Focus on the failing test.", isFirstVisit: true }),
+    );
+
+    expect(result).toMatch(/^Use YAML frontmatter/);
+    expect(result).toContain("Write code");
+    expect(result).toContain("## Task\nFix the bug");
+    expect(result).toContain("## Moderator Instruction");
+    expect(result).toContain("Focus on the failing test.");
+  });
+
+  test("re-entry uses continuation prompt with edge instruction", () => {
+    const ctx = makeCtx({
+      isFirstVisit: false,
+      edgePrompt: "The reviewer rejected your work. Fix the issues.",
+      steps: [
+        {
+          role: "developer",
+          output: { summary: "Initial fix" },
+          agent: "uwf-hermes",
+          detail: "detail-1",
+          edgePrompt: "Implement the fix.",
+        },
+        {
+          role: "reviewer",
+          output: { approved: false },
+          agent: "uwf-hermes",
+          detail: "detail-2",
+          edgePrompt: "Review the code.",
+        },
+      ],
+    });
+
+    const result = buildHermesPrompt(ctx);
+
+    expect(result).not.toContain("## Task");
+    expect(result).toContain("## What Happened Since Your Last Turn");
+    expect(result).toContain("## Moderator Instruction");
+    expect(result).toContain("The reviewer rejected your work.");
+  });
+
+  test("forced first visit via isFirstVisit uses initial prompt even when role appears in history", () => {
+    const result = buildHermesPrompt(
+      makeCtx({
+        isFirstVisit: true,
+        steps: [
+          {
+            role: "developer",
+            output: { done: true },
+            agent: "uwf-hermes",
+            detail: "detail-1",
+            edgePrompt: "First attempt.",
+          },
+        ],
+        edgePrompt: "Retry with a fresh approach.",
+      }),
+    );
+
+    expect(result).toContain("## Task");
+    expect(result).toContain("Retry with a fresh approach.");
+    expect(result).not.toContain("## What Happened Since Your Last Turn");
+  });
+});

packages/workflow-agent-hermes/__tests__/resume-e2e.test.ts

@@ -0,0 +1,57 @@
+import { afterEach, describe, expect, it } from "bun:test";
+
+import { HermesAcpClient } from "../src/acp-client.js";
+
+/**
+ * E2E test for cross-process session resume.
+ *
+ * Simulates the workflow re-entry scenario:
+ * 1. Client A: connect → prompt → close (developer first run)
+ * 2. Client B: resume(sessionId) → prompt (developer re-entry after reviewer reject)
+ *
+ * This is what happens when uwf thread step spawns uwf-hermes twice for the same role.
+ */
+describe("HermesAcpClient cross-process resume", () => {
+  const clients: HermesAcpClient[] = [];
+
+  afterEach(async () => {
+    for (const c of clients) {
+      await c.close();
+    }
+    clients.length = 0;
+  });
+
+  // TODO(#435): flaky — depends on live LLM; mock or move to integration suite
+  it.skip(
+    "resume() after close — second prompt returns non-empty text",
+    async () => {
+      // --- Client A: first run ---
+      const clientA = new HermesAcpClient();
+      clients.push(clientA);
+
+      await clientA.connect(process.cwd());
+      const first = await clientA.prompt(
+        "Remember the secret code: WATERMELON. Reply with exactly: ACKNOWLEDGED",
+      );
+      expect(first.text.length).toBeGreaterThan(0);
+      const sessionId = first.sessionId;
+
+      // Close client A (simulates uwf-hermes process exit)
+      await clientA.close();
+
+      // --- Client B: resume (simulates re-entry) ---
+      const clientB = new HermesAcpClient();
+      clients.push(clientB);
+
+      await clientB.resume(sessionId, process.cwd());
+      const second = await clientB.prompt(
+        "What was the secret code I told you earlier? Reply with just the code word.",
+      );
+
+      // The critical assertion: resumed session produces non-empty output
+      expect(second.text.length).toBeGreaterThan(0);
+      expect(second.sessionId).toBe(sessionId);
+    },
+    { timeout: 3 * 60 * 1000 },
+  );
+});

packages/workflow-agent-hermes/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@uncaged/workflow-agent-hermes",
-  "version": "0.1.0",
+  "version": "0.5.0",
   "files": [
     "src",
     "dist",
@@ -21,8 +21,10 @@
     "test": "bun test"
   },
   "dependencies": {
-    "@uncaged/json-cas": "^0.3.0",
-    "@uncaged/workflow-agent-kit": "workspace:^"
+    "@uncaged/json-cas": "^0.4.0",
+    "@uncaged/workflow-agent-kit": "workspace:^",
+    "@uncaged/workflow-protocol": "workspace:^",
+    "@uncaged/workflow-util": "workspace:^"
   },
   "devDependencies": {
     "typescript": "^5.8.3"

packages/workflow-agent-hermes/src/acp-client.ts

@@ -0,0 +1,395 @@
+import type { ChildProcess } from "node:child_process";
+import { spawn } from "node:child_process";
+import { createInterface } from "node:readline";
+
+import type { HermesSessionMessage } from "./types.js";
+
+const HERMES_COMMAND = "hermes";
+const PROTOCOL_VERSION = 1;
+
+type JsonRpcResponse = {
+  jsonrpc: "2.0";
+  id: number;
+  result?: unknown;
+  error?: { code: number; message: string };
+};
+
+type PendingRequest = {
+  resolve: (value: JsonRpcResponse) => void;
+  reject: (reason: Error) => void;
+};
+
+/** Tracks in-flight tool calls so we can build complete messages when they finish. */
+type PendingToolCall = {
+  name: string;
+  args: string;
+};
+
+export type AcpPromptResult = {
+  text: string;
+  sessionId: string;
+  messages: HermesSessionMessage[];
+};
+
+export class HermesAcpClient {
+  private process: ChildProcess | null = null;
+  private nextId = 1;
+  private sessionId: string | null = null;
+  private stderrBuffer = "";
+  private pending = new Map<number, PendingRequest>();
+
+  // Message collection state
+  private messageChunks: string[] = [];
+  private reasoningChunks: string[] = [];
+  private pendingTools = new Map<string, PendingToolCall>();
+  messages: HermesSessionMessage[] = [];
+
+  /** Spawn hermes acp, initialize, create session */
+  async connect(cwd: string): Promise<string> {
+    await this.ensureProcess();
+    await this.initialize();
+
+    const sessionResponse = (await this.sendRequest("session/new", {
+      cwd,
+      mcpServers: [],
+    })) as { result: { sessionId: string } };
+
+    const sessionId = sessionResponse.result?.sessionId;
+    if (typeof sessionId !== "string" || sessionId === "") {
+      throw new Error(`session/new did not return a sessionId: ${JSON.stringify(sessionResponse)}`);
+    }
+
+    this.sessionId = sessionId;
+    return sessionId;
+  }
+
+  /** Spawn hermes acp, initialize, resume an existing session */
+  async resume(sessionId: string, cwd: string): Promise<string> {
+    await this.ensureProcess();
+    await this.initialize();
+
+    const response = await this.sendRequest("session/resume", {
+      cwd,
+      sessionId,
+      mcpServers: [],
+    });
+
+    if ((response as { error?: unknown }).error !== undefined) {
+      throw new Error(
+        `session/resume failed: ${JSON.stringify((response as { error: unknown }).error)}`,
+      );
+    }
+
+    this.sessionId = sessionId;
+    return sessionId;
+  }
+
+  /** Send prompt and collect full response text + structured messages. */
+  async prompt(text: string): Promise<AcpPromptResult> {
+    if (this.sessionId === null) {
+      throw new Error("Not connected — call connect() first");
+    }
+
+    this.messageChunks = [];
+    this.reasoningChunks = [];
+
+    const response = await this.sendRequest("session/prompt", {
+      sessionId: this.sessionId,
+      prompt: [{ type: "text", text }],
+    });
+
+    if ((response as { error?: unknown }).error !== undefined) {
+      throw new Error(
+        `session/prompt failed: ${JSON.stringify((response as { error: unknown }).error)}`,
+      );
+    }
+
+    // Flush any trailing assistant text that wasn't followed by a tool call.
+    this.flushAssistantMessage();
+
+    // Extract the final assistant text from collected messages.
+    let finalText = "";
+    for (let i = this.messages.length - 1; i >= 0; i--) {
+      const msg = this.messages[i];
+      if (
+        msg !== undefined &&
+        msg.role === "assistant" &&
+        msg.content !== null &&
+        msg.content.trim() !== ""
+      ) {
+        finalText = msg.content;
+        break;
+      }
+    }
+
+    return {
+      text: finalText,
+      sessionId: this.sessionId,
+      messages: this.messages,
+    };
+  }
+
+  /** Close the connection */
+  async close(): Promise<void> {
+    if (this.process === null) {
+      return;
+    }
+    this.sessionId = null;
+    this.process.stdin?.end();
+    const proc = this.process;
+    await new Promise<void>((resolve) => {
+      proc.on("close", () => resolve());
+      setTimeout(resolve, 5000);
+    });
+    this.process = null;
+  }
+
+  // ---- JSON-RPC transport ----
+
+  private sendRequest(
+    method: string,
+    params: Record<string, unknown>,
+    timeoutMs = 10 * 60 * 1000,
+  ): Promise<JsonRpcResponse> {
+    const id = this.nextId++;
+    return new Promise<JsonRpcResponse>((resolve, reject) => {
+      const timer = setTimeout(() => {
+        this.pending.delete(id);
+        reject(new Error(`Timeout waiting for response to ${method} (id=${id})`));
+      }, timeoutMs);
+
+      this.pending.set(id, {
+        resolve: (value) => {
+          clearTimeout(timer);
+          resolve(value);
+        },
+        reject: (err) => {
+          clearTimeout(timer);
+          reject(err);
+        },
+      });
+
+      this.writeLine(JSON.stringify({ jsonrpc: "2.0", id, method, params }));
+    });
+  }
+
+  private sendNotification(method: string, params?: Record<string, unknown>): void {
+    const message: Record<string, unknown> = { jsonrpc: "2.0", method };
+    if (params !== undefined) {
+      message.params = params;
+    }
+    this.writeLine(JSON.stringify(message));
+  }
+
+  private writeLine(line: string): void {
+    if (this.process?.stdin === null || this.process?.stdin === undefined) {
+      throw new Error("Cannot write: hermes acp process stdin not available");
+    }
+    this.process.stdin.write(`${line}\n`);
+  }
+
+  private handleLine(line: string): void {
+    if (line === "") {
+      return;
+    }
+
+    let parsed: unknown;
+    try {
+      parsed = JSON.parse(line);
+    } catch {
+      return;
+    }
+
+    const msg = parsed as Record<string, unknown>;
+
+    const hasId = "id" in msg && msg.id !== undefined && msg.id !== null;
+    const hasMethod = typeof msg.method === "string";
+
+    // JSON-RPC response to one of our requests (has "id" but no "method")
+    if (hasId && !hasMethod) {
+      const response = msg as unknown as JsonRpcResponse;
+      const handler = this.pending.get(response.id);
+      if (handler !== undefined) {
+        this.pending.delete(response.id);
+        handler.resolve(response);
+      }
+      return;
+    }
+
+    // Server-initiated JSON-RPC request: session/request_permission (has "id" + "method")
+    if (msg.method === "session/request_permission" && hasId) {
+      const params = msg.params as Record<string, unknown> | undefined;
+      const options = (params?.options ?? []) as Array<{ optionId?: string }>;
+      const firstOptionId = options[0]?.optionId ?? "";
+      this.writeLine(
+        JSON.stringify({
+          jsonrpc: "2.0",
+          id: msg.id,
+          result: { outcome: { outcome: "selected", optionId: firstOptionId } },
+        }),
+      );
+      return;
+    }
+
+    // JSON-RPC notification — session/update (no "id")
+    if (msg.method === "session/update") {
+      const params = msg.params as Record<string, unknown> | undefined;
+      const update = params?.update as Record<string, unknown> | undefined;
+      if (update !== undefined) {
+        this.handleSessionUpdate(update);
+      }
+      return;
+    }
+  }
+
+  // ---- Session update → structured messages ----
+
+  private handleSessionUpdate(update: Record<string, unknown>): void {
+    switch (update.sessionUpdate as string) {
+      case "agent_message_chunk":
+        this.handleAgentMessageChunk(update);
+        break;
+      case "agent_thought_chunk":
+        this.handleAgentThoughtChunk(update);
+        break;
+      case "tool_call":
+        this.handleToolCall(update);
+        break;
+      case "tool_call_update":
+        this.handleToolCallUpdate(update);
+        break;
+      default:
+        break;
+    }
+  }
+
+  private handleAgentMessageChunk(update: Record<string, unknown>): void {
+    const content = update.content as { type?: string; text?: string } | undefined;
+    if (content?.type === "text" && typeof content.text === "string") {
+      this.messageChunks.push(content.text);
+    }
+  }
+
+  private handleAgentThoughtChunk(update: Record<string, unknown>): void {
+    const content = update.content as { type?: string; text?: string } | undefined;
+    if (content?.type === "text" && typeof content.text === "string") {
+      this.reasoningChunks.push(content.text);
+    }
+  }
+
+  private handleToolCall(update: Record<string, unknown>): void {
+    const title = (update.title as string) ?? "";
+    const rawInput = update.rawInput;
+    const args = rawInput !== undefined && rawInput !== null ? JSON.stringify(rawInput) : "";
+    const toolCallId = update.toolCallId as string;
+    this.pendingTools.set(toolCallId, { name: title, args });
+    this.flushAssistantMessage();
+  }
+
+  private handleToolCallUpdate(update: Record<string, unknown>): void {
+    const status = update.status as string | undefined;
+    if (status !== "completed" && status !== "failed") return;
+    const toolCallId = update.toolCallId as string;
+    const pending = this.pendingTools.get(toolCallId);
+    const toolName = pending?.name ?? toolCallId;
+    const rawOutput = update.rawOutput;
+    const outputStr =
+      rawOutput !== undefined && rawOutput !== null
+        ? typeof rawOutput === "string"
+          ? rawOutput
+          : JSON.stringify(rawOutput)
+        : "";
+    this.messages.push({
+      role: "assistant",
+      content: null,
+      reasoning: null,
+      tool_calls: [{ function: { name: toolName, arguments: pending?.args ?? "" } }],
+    });
+    this.messages.push({
+      role: "tool",
+      content: outputStr,
+      reasoning: null,
+      tool_calls: null,
+    });
+    this.pendingTools.delete(toolCallId);
+  }
+
+  /** Flush any accumulated text/reasoning into an assistant message. */
+  private flushAssistantMessage(): void {
+    const text = this.messageChunks.join("");
+    const reasoning = this.reasoningChunks.join("");
+    if (text !== "" || reasoning !== "") {
+      this.messages.push({
+        role: "assistant",
+        content: text || null,
+        reasoning: reasoning || null,
+        tool_calls: null,
+      });
+    }
+    this.messageChunks = [];
+    this.reasoningChunks = [];
+  }
+
+  private rejectAll(err: Error): void {
+    for (const handler of this.pending.values()) {
+      handler.reject(err);
+    }
+    this.pending.clear();
+  }
+
+  private async ensureProcess(): Promise<void> {
+    if (this.process !== null) {
+      return;
+    }
+
+    const child = spawn(HERMES_COMMAND, ["acp"], {
+      env: process.env,
+      shell: false,
+      stdio: ["pipe", "pipe", "pipe"],
+    });
+
+    this.process = child;
+
+    child.stderr?.on("data", (chunk: Buffer) => {
+      this.stderrBuffer += chunk.toString();
+    });
+
+    child.on("error", (cause) => {
+      const message = cause instanceof Error ? cause.message : String(cause);
+      this.rejectAll(new Error(`hermes acp spawn failed: ${message}`));
+    });
+
+    child.on("close", (code) => {
+      if (code !== 0 && this.pending.size > 0) {
+        const detail = this.stderrBuffer.trim() !== "" ? ` stderr=${this.stderrBuffer.trim()}` : "";
+        this.rejectAll(
+          new Error(`hermes acp exited unexpectedly with code ${code ?? "null"}${detail}`),
+        );
+      }
+    });
+
+    if (child.stdout === null) {
+      throw new Error("hermes acp process stdout is not available");
+    }
+    const rl = createInterface({ input: child.stdout });
+    rl.on("line", (line) => {
+      this.handleLine(line.trim());
+    });
+  }
+
+  private async initialize(): Promise<void> {
+    const initResponse = await this.sendRequest("initialize", {
+      protocolVersion: PROTOCOL_VERSION,
+      clientInfo: { name: "uwf", version: "0.1.0" },
+      capabilities: {},
+    });
+
+    if ((initResponse as { error?: unknown }).error !== undefined) {
+      throw new Error(
+        `initialize failed: ${JSON.stringify((initResponse as { error: unknown }).error)}`,
+      );
+    }
+
+    this.sendNotification("initialized");
+  }
+}

packages/workflow-agent-hermes/src/hermes.ts

@@ -1,16 +1,18 @@
-import { spawn } from "node:child_process";
-
-import { type AgentContext, type AgentRunResult, createAgent } from "@uncaged/workflow-agent-kit";
-
+import type { Store } from "@uncaged/json-cas";
 import {
-  loadHermesSession,
-  parseSessionIdFromStdout,
-  storeHermesRawOutput,
-  storeHermesSessionDetail,
-} from "./session-detail.js";
+  type AgentContext,
+  type AgentRunResult,
+  buildContinuationPrompt,
+  buildRolePrompt,
+  createAgent,
+} from "@uncaged/workflow-agent-kit";
+import { createLogger } from "@uncaged/workflow-util";
 
-const HERMES_COMMAND = "hermes";
-const HERMES_MAX_TURNS = 90;
+import { HermesAcpClient } from "./acp-client.js";
+import { getCachedSessionId, isResumeDisabled, setCachedSessionId } from "./session-cache.js";
+import { storeHermesSessionDetail } from "./session-detail.js";
+
+const log = createLogger({ sink: { kind: "stderr" } });
 
 function buildHistorySummary(steps: AgentContext["steps"]): string {
   if (steps.length === 0) {
@@ -31,87 +33,157 @@ function buildHistorySummary(steps: AgentContext["steps"]): string {
   return lines.join("\n");
 }
 
-/** Assemble system prompt, task, and prior step outputs for Hermes. */
-export function buildHermesPrompt(ctx: AgentContext): string {
+function buildInitialPrompt(ctx: AgentContext): string {
   const roleDef = ctx.workflow.roles[ctx.role];
-  const systemPrompt = roleDef?.systemPrompt ?? "";
+  const rolePrompt = roleDef !== undefined ? buildRolePrompt(roleDef) : "";
   const parts: string[] = [];
-  if (ctx.outputFormatInstruction !== undefined && ctx.outputFormatInstruction !== "") {
+  if (ctx.outputFormatInstruction !== "") {
     parts.push(ctx.outputFormatInstruction, "");
   }
-  parts.push(systemPrompt, "", "## Task", ctx.start.prompt);
+  parts.push(rolePrompt, "", "## Task", ctx.start.prompt);
   const historyBlock = buildHistorySummary(ctx.steps);
   if (historyBlock !== "") {
     parts.push("", historyBlock);
   }
+  parts.push("", "## Moderator Instruction", "", ctx.edgePrompt);
   return parts.join("\n");
 }
 
-function spawnHermesChat(prompt: string): Promise<{ stdout: string; stderr: string }> {
-  return new Promise((resolve, reject) => {
-    const args = [
-      "chat",
-      "-q",
-      prompt,
-      "--yolo",
-      "--max-turns",
-      String(HERMES_MAX_TURNS),
-      "--quiet",
-    ];
-    const child = spawn(HERMES_COMMAND, args, {
-      env: process.env,
-      shell: false,
-      stdio: ["ignore", "pipe", "pipe"],
-    });
+/** Assemble system prompt, task, and prior step outputs for Hermes. */
+export function buildHermesPrompt(ctx: AgentContext): string {
+  if (!ctx.isFirstVisit) {
+    const parts: string[] = [];
+    if (ctx.outputFormatInstruction !== "") {
+      parts.push(ctx.outputFormatInstruction, "");
+    }
+    parts.push(buildContinuationPrompt(ctx.steps, ctx.role, ctx.edgePrompt));
+    return parts.join("\n");
+  }
 
-    let stdout = "";
-    let stderr = "";
-    child.stdout?.on("data", (chunk: Buffer) => {
-      stdout += chunk.toString();
-    });
-    child.stderr?.on("data", (chunk: Buffer) => {
-      stderr += chunk.toString();
-    });
-
-    child.on("error", (cause) => {
-      const message = cause instanceof Error ? cause.message : String(cause);
-      reject(new Error(`hermes spawn failed: ${message}`));
-    });
-
-    child.on("close", (code) => {
-      if (code === 0) {
-        resolve({ stdout, stderr });
-        return;
-      }
-      const detail = stderr.trim() !== "" ? ` stderr=${stderr.trim()}` : "";
-      reject(new Error(`hermes exited with code ${code ?? "null"}${detail}`));
-    });
-  });
+  return buildInitialPrompt(ctx);
 }
 
-async function runHermes(ctx: AgentContext): Promise<AgentRunResult> {
-  const fullPrompt = buildHermesPrompt(ctx);
-  const { stdout, stderr } = await spawnHermesChat(fullPrompt);
-  const { store } = ctx;
+async function storePromptResult(
+  store: Store,
+  sessionId: string,
+  messages: Awaited<ReturnType<HermesAcpClient["prompt"]>>["messages"],
+): Promise<{ detailHash: string }> {
+  const session = {
+    session_id: sessionId,
+    model: "",
+    session_start: new Date().toISOString(),
+    messages,
+  };
+  return storeHermesSessionDetail(store, session);
+}
 
-  // --quiet mode: session_id may be on stdout or stderr
-  const sessionId = parseSessionIdFromStdout(stderr) ?? parseSessionIdFromStdout(stdout);
-  if (sessionId !== null) {
-    const session = await loadHermesSession(sessionId);
-    if (session !== null) {
-      const { detailHash, output } = await storeHermesSessionDetail(store, session);
-      return { output, detailHash };
+type PromptAttempt = {
+  useContinuation: boolean;
+  resumed: boolean;
+};
+
+async function prepareSession(
+  client: HermesAcpClient,
+  ctx: AgentContext,
+  cwd: string,
+): Promise<PromptAttempt> {
+  if (ctx.isFirstVisit || isResumeDisabled()) {
+    await client.connect(cwd);
+    return { useContinuation: false, resumed: false };
+  }
+
+  const cachedSessionId = await getCachedSessionId(ctx.threadId, ctx.role);
+  if (cachedSessionId === null) {
+    log("6RWK3N8Q", `no cached session for ${ctx.threadId}:${ctx.role}, starting new session`);
+    await client.connect(cwd);
+    return { useContinuation: false, resumed: false };
+  }
+
+  try {
+    await client.resume(cachedSessionId, cwd);
+    log("9MHT4V2P", `resumed hermes session ${cachedSessionId} for ${ctx.threadId}:${ctx.role}`);
+    return { useContinuation: true, resumed: true };
+  } catch (error) {
+    const message = error instanceof Error ? error.message : String(error);
+    log("3XPN7K4W", `session resume failed, falling back to new session: ${message}`);
+    await client.close();
+    await client.connect(cwd);
+    return { useContinuation: false, resumed: false };
+  }
+}
+
+/**
+ * Agent CLI factory: parses argv, runs Hermes, extracts output, writes StepNode.
+ *
+ * A single ACP client is shared across run() and continue() calls so that
+ * frontmatter retry loops keep the same Hermes session context.  The client
+ * is closed once the agent process exits (via process.on("exit")).
+ */
+export function createHermesAgent(): () => Promise<void> {
+  const client = new HermesAcpClient();
+
+  // Ensure cleanup regardless of how the process exits.
+  process.on("exit", () => {
+    void client.close();
+  });
+
+  async function runPrompt(ctx: AgentContext, useContinuation: boolean): Promise<AgentRunResult> {
+    const effectiveCtx = useContinuation ? ctx : { ...ctx, isFirstVisit: true };
+    const fullPrompt = buildHermesPrompt(effectiveCtx);
+    const { text, sessionId, messages } = await client.prompt(fullPrompt);
+    const { detailHash } = await storePromptResult(ctx.store, sessionId, messages);
+
+    if (!isResumeDisabled()) {
+      await setCachedSessionId(ctx.threadId, ctx.role, sessionId);
+    }
+
+    return { output: text, detailHash, sessionId };
+  }
+
+  async function runHermes(ctx: AgentContext): Promise<AgentRunResult> {
+    const cwd = process.cwd();
+    const attempt = await prepareSession(client, ctx, cwd);
+
+    try {
+      return await runPrompt(ctx, attempt.useContinuation);
+    } catch (error) {
+      if (!attempt.resumed) {
+        throw error;
+      }
+
+      const message = error instanceof Error ? error.message : String(error);
+      log("8FQW2R6N", `continuation prompt failed, retrying with initial prompt: ${message}`);
+      await client.close();
+      await client.connect(cwd);
+      return runPrompt(ctx, false);
     }
   }
 
-  const detailHash = await storeHermesRawOutput(store, stdout);
-  return { output: stdout, detailHash };
-}
+  async function continueHermes(
+    _sessionId: string,
+    message: string,
+    store: Store,
+  ): Promise<AgentRunResult> {
+    // Client is already connected from runHermes — same ACP session,
+    // so the agent sees the full conversation history (crucial for retries).
+    const { text, sessionId, messages } = await client.prompt(message);
+    const { detailHash } = await storePromptResult(store, sessionId, messages);
+    return { output: text, detailHash, sessionId };
+  }
 
-/** Agent CLI factory: parses argv, runs Hermes, extracts output, writes StepNode. */
-export function createHermesAgent(): () => Promise<void> {
-  return createAgent({
+  const agentMain = createAgent({
     name: "hermes",
     run: runHermes,
+    continue: continueHermes,
   });
+
+  // Wrap to ensure ACP client is closed after agent completes,
+  // so the hermes subprocess exits and bun can terminate.
+  return async () => {
+    try {
+      await agentMain();
+    } finally {
+      await client.close();
+    }
+  };
 }

packages/workflow-agent-hermes/src/index.ts

@@ -1 +1,2 @@
+export { HermesAcpClient } from "./acp-client.js";
 export { buildHermesPrompt, createHermesAgent } from "./hermes.js";

packages/workflow-agent-hermes/src/session-cache.ts

@@ -0,0 +1,34 @@
+// Re-export session cache from the shared agent-kit package with agent name injected.
+
+import {
+  getCachedSessionId as getCachedSessionIdBase,
+  setCachedSessionId as setCachedSessionIdBase,
+} from "@uncaged/workflow-agent-kit";
+import type { ThreadId } from "@uncaged/workflow-protocol";
+
+export async function getCachedSessionId(threadId: ThreadId, role: string): Promise<string | null> {
+  return getCachedSessionIdBase("hermes", threadId, role);
+}
+
+export async function setCachedSessionId(
+  threadId: ThreadId,
+  role: string,
+  sessionId: string,
+): Promise<void> {
+  return setCachedSessionIdBase("hermes", threadId, role, sessionId);
+}
+
+export function isResumeDisabled(): boolean {
+  // Hermes ACP session/resume is broken: _restore fails for custom providers
+  // because resolve_runtime_provider("custom") throws and base_url/api_mode
+  // are lost in the fallback path.  Resume silently creates a new session
+  // (different sessionId, no history), causing empty-text responses.
+  // See: https://github.com/NousResearch/hermes-agent/issues/13489
+  // Disable by default until upstream fixes the bug.  Set UWF_HERMES_RESUME=1
+  // to opt back in.
+  const enableFlag = process.env.UWF_HERMES_RESUME;
+  if (enableFlag === "1" || enableFlag === "true") {
+    return false;
+  }
+  return true;
+}

packages/workflow-agent-kit/README.md

@@ -0,0 +1,182 @@
+# @uncaged/workflow-agent-kit
+
+Agent framework — `createAgent` factory, context builder, frontmatter fast-path, and LLM extract pipeline.
+
+## Overview
+
+Layer 2 agent framework. Provides the standard entrypoint for all agent CLIs: parse `<thread-id> <role>` from argv, load thread/workflow context from CAS, invoke the agent's `run`/`continue` functions, validate output via frontmatter fast-path or LLM extract, and write a `StepNodePayload` to CAS.
+
+Also exports prompt builders, config/storage helpers, and session ID caching for multi-turn agents.
+
+**Dependencies:** `@uncaged/json-cas`, `@uncaged/json-cas-fs`, `@uncaged/workflow-protocol`, `@uncaged/workflow-util`, `dotenv`, `yaml`
+
+## Installation
+
+```bash
+bun add @uncaged/workflow-agent-kit
+```
+
+## API
+
+All exports come from `src/index.ts`.
+
+### Agent factory
+
+```typescript
+function createAgent(options: AgentOptions): () => Promise<void>
+
+type AgentOptions = {
+  name: string;
+  run: AgentRunFn;
+  continue: AgentContinueFn;
+};
+
+type AgentRunFn = (ctx: AgentContext) => Promise<AgentRunResult>;
+type AgentContinueFn = (
+  sessionId: string,
+  message: string,
+  store: AgentContext["store"],
+) => Promise<AgentRunResult>;
+
+type AgentRunResult = {
+  output: string;
+  detailHash: string;
+  sessionId: string;
+};
+```
+
+Agent CLIs call `createAgent(...)` and invoke the returned function as `main()`.
+
+### Context
+
+```typescript
+function buildContext(threadId: ThreadId, role: string): Promise<AgentContext>
+function buildContextWithMeta(
+  threadId: ThreadId,
+  role: string,
+): Promise<AgentContext & { meta: BuildContextMeta }>
+
+type AgentContext = ModeratorContext & {
+  threadId: ThreadId;
+  role: string;
+  store: Store;
+  workflow: WorkflowPayload;
+  outputFormatInstruction: string;
+  edgePrompt: string;
+  isFirstVisit: boolean;
+};
+
+type BuildContextMeta = {
+  storageRoot: string;
+  store: Store;
+  schemas: AgentStore["schemas"];
+  headHash: CasRef;
+  chain: ChainState;
+};
+```
+
+Requires `UWF_EDGE_PROMPT` in the environment (set by `uwf thread step`).
+
+### Prompt builders
+
+```typescript
+function buildRolePrompt(role: RoleDefinition): string
+function buildOutputFormatInstruction(schema: JSONSchema): string
+function buildContinuationPrompt(
+  ctx: AgentContext,
+  priorOutput: string,
+  instruction: string,
+): string
+```
+
+### Extract pipeline
+
+```typescript
+function resolveExtractModelAlias(config: WorkflowConfig): ModelAlias
+function resolveModel(config: WorkflowConfig, alias: ModelAlias): ResolvedLlmProvider
+function extract(
+  rawOutput: string,
+  outputSchema: CasRef,
+  config: WorkflowConfig,
+): Promise<ExtractResult>
+
+type ResolvedLlmProvider = { baseUrl: string; apiKey: string; model: string };
+type ExtractResult = { value: unknown; hash: CasRef };
+```
+
+### Frontmatter fast-path
+
+```typescript
+function tryFrontmatterFastPath(
+  rawOutput: string,
+  outputSchema: CasRef,
+  store: Store,
+): Promise<FrontmatterFastPathResult | null>
+
+type FrontmatterFastPathResult = { body: string; outputHash: CasRef };
+```
+
+### Session cache
+
+```typescript
+function getCachedSessionId(threadId: ThreadId, role: string): Promise<string | null>
+function setCachedSessionId(
+  threadId: ThreadId,
+  role: string,
+  sessionId: string,
+): Promise<void>
+```
+
+### Config and storage
+
+```typescript
+function getConfigPath(storageRoot: string): string
+function getEnvPath(storageRoot: string): string
+function resolveStorageRoot(): string
+function loadWorkflowConfig(storageRoot: string): Promise<WorkflowConfig>
+```
+
+## Usage
+
+```typescript
+import { createAgent, buildRolePrompt } from "@uncaged/workflow-agent-kit";
+import type { AgentContext, AgentRunResult } from "@uncaged/workflow-agent-kit";
+
+async function run(ctx: AgentContext): Promise<AgentRunResult> {
+  const prompt = buildRolePrompt(ctx.workflow.roles[ctx.role]!);
+  // ... spawn external process, capture output ...
+  return { output: markdown, detailHash: "...", sessionId: "..." };
+}
+
+async function continueSession(
+  sessionId: string,
+  message: string,
+): Promise<AgentRunResult> {
+  // ... continue multi-turn session ...
+  return { output: markdown, detailHash: "...", sessionId };
+}
+
+export const main = createAgent({ name: "my-agent", run, continue: continueSession });
+```
+
+## Internal Structure
+
+```
+src/
+├── index.ts
+├── run.ts                         createAgent entrypoint
+├── context.ts                     Thread chain walk, AgentContext builder
+├── extract.ts                     LLM structured extract fallback
+├── frontmatter.ts                 Frontmatter fast-path validation
+├── build-role-prompt.ts           Role definition → prompt text
+├── build-output-format-instruction.ts
+├── build-continuation-prompt.ts
+├── session-cache.ts               Per-thread/session ID persistence
+├── storage.ts                     CAS store, config, threads index
+├── schemas.ts                     Agent CAS schema registration
+└── types.ts                       AgentContext, AgentOptions, etc.
+```
+
+## Configuration
+
+Reads `config.yaml` and `.env` from the workflow storage root (`~/.uncaged/workflow` by default). See `@uncaged/workflow-protocol` for `WorkflowConfig` shape. Set via `uwf setup`.