refactor: rename RoleDefinition fields for clarity

- identity → goal - prepare → capabilities (string[]) - execute → procedure - report → output - outputSchema → meta Fixes #364
Merge pull request 'feat: Role 四段式描述 (identity/prepare/execute/report)' (#361 ) from feat/359-role-four-phase into main
2026-05-22 00:46:06 +00:00 · 2026-05-21 03:11:00 +00:00 · 2026-05-21 03:07:56 +00:00 · 2026-05-21 02:56:11 +00:00 · 2026-05-21 02:31:56 +00:00 · 2026-05-21 01:41:20 +00:00
511 changed files with 13301 additions and 8614 deletions
@@ -1,35 +0,0 @@
-# Uncaged Workflow Architecture
-
-Uncaged Workflow is a monorepo implementing a workflow engine that executes single-file ESM bundles. Each workflow is identified by an XXH64 hash (Crockford Base32); execution state is stored in a content-addressable store (CAS) as immutable Merkle nodes. Agents are pluggable — the same workflow definition runs with Cursor, Hermes, a raw LLM, or a ReAct loop.
-
-## Core Concepts
-
-| Card | Description |
-|------|-------------|
-| [Bundle](./bundle.md) | A single-file `.esm.js` module with an XXH64 hash identity, stored in `~/.uncaged/workflow/bundles/` |
-| [Thread](./thread.md) | A single execution instance of a workflow, identified by a ULID, with CAS-linked state nodes |
-| [CAS](./cas.md) | The content-addressable store that holds all immutable blobs — content, start nodes, and state nodes |
-| [Registry](./registry.md) | `workflow.yaml` — maps workflow names to current and historical bundle hashes |
-
-## Execution
-
-| Card | Description |
-|------|-------------|
-| [Engine](./engine.md) | The three-phase loop that drives the workflow `AsyncGenerator` and writes each step to CAS |
-| [Role](./role.md) | A named actor defined as pure data (`RoleDefinition`) — description, system prompt, and Zod schema |
-| [Agent Binding](./agent-binding.md) | The runtime binding that connects a role to a concrete agent implementation via `AdapterFn` |
-| [Reactor](./reactor.md) | The ReAct loop abstraction for LLM function-calling, used by both the extract phase and agent adapters |
-
-## Tooling
-
-| Card | Description |
-|------|-------------|
-| [CLI](./cli.md) | The `uncaged-workflow` command-line tool for managing workflows, threads, and CAS |
-| [Dashboard](./dashboard.md) | A private React app for inspecting threads, workflows, and live execution via the gateway |
-| [Package Map](./package-map.md) | All packages in the monorepo with their layer positions and dependency graph |
-
-## Authoring
-
-| Card | Description |
-|------|-------------|
-| [Workflow Templates](./workflow-templates.md) | The `solve-issue` and `develop` reference templates and how to author custom workflows |
@@ -1,104 +0,0 @@
-# Agent Binding
-
-> The runtime connection between a workflow's role definitions and a concrete agent implementation, expressed as an `AdapterBinding` passed to `createWorkflow`.
-
-## Overview
-
-Agent binding is how a workflow author specifies which agent executes each role. Roles are pure data (see [Role](./role.md)); the binding supplies the execution strategy. The same `WorkflowDefinition` can be run with different agents by changing the `AdapterBinding` — useful for testing, cost optimization, or environment-specific deployment.
-
-An `AdapterFn` receives a role's `systemPrompt` and Zod `schema`, and returns a `RoleFn` — a function that takes `ThreadContext` and `WorkflowRuntime` and returns `RoleResult<T>`. The adapter is responsible for producing typed structured output directly; there is no separate extract phase when using adapters.
-
-## Key Types
-
-```typescript
-// The core adapter interface
-type AdapterFn = <T>(prompt: string, schema: z.ZodType<T>) => RoleFn<T>;
-
-type RoleFn<T> = (ctx: ThreadContext, runtime: WorkflowRuntime) => Promise<RoleResult<T>>;
-
-type RoleResult<T> = { meta: T; childThread: string | null };
-
-// The binding passed to createWorkflow
-type AdapterBinding = {
-  adapter: AdapterFn;
-  overrides: Partial<Record<string, AdapterFn>> | null;
-};
-```
-
-`overrides` allows per-role adapters — for example, using Cursor for one role and an LLM for another within the same workflow.
-
-## AgentFn (Legacy / Low-level)
-
-Below the adapter layer, the original `AgentFn` type still exists for agent implementations that produce raw strings rather than structured output:
-
-```typescript
-type AgentFn<Opt = void> = Opt extends void
-  ? (ctx: ThreadContext) => Promise<string>
-  : (ctx: ThreadContext, options: Opt) => Promise<string>;
-```
-
-The `createAgentAdapter` utility in `@uncaged/workflow-util-agent` wraps an `AgentFn` into an `AdapterFn` by composing it with extraction logic.
-
-## Concrete Implementations
-
-| Package | Export | Agent |
-|---------|--------|-------|
-| `@uncaged/workflow-agent-cursor` | `createCursorAgent` | Runs `cursor` CLI non-interactively in a workspace directory |
-| `@uncaged/workflow-agent-hermes` | `createHermesAgent` | Runs `hermes chat` with `--yolo --quiet` (Nerve-style argv) |
-| `@uncaged/workflow-agent-llm` | `createLlmAdapter` | Direct LLM completion via the OpenAI-compatible chat endpoint |
-| `@uncaged/workflow-agent-react` | `createReactAdapter` | ReAct loop with file and shell tools (read, write, patch, exec) |
-
-All four return an `AdapterFn` suitable for use in `AdapterBinding.adapter`.
-
-## workflow-util-agent
-
-`@uncaged/workflow-util-agent` provides two helpers shared by adapter implementations:
-
- **`buildThreadInput(ctx)`** — constructs the user-message string from thread context (task, previous steps, tool hints). Used by all CLI-based agents.
- **`spawnCli(command, args, opts)`** — spawns an external process (e.g., `cursor`, `hermes`) and captures stdout, with optional timeout.
- **`createAgentAdapter(agentFn, optionsFn)`** — wraps an `AgentFn<Opt>` into an `AdapterFn`, handling the options extraction step.
-
-## Cursor Agent
-
-`createCursorAgent(config)` invokes the `cursor` CLI binary:
-
-```
-cursor -p <fullPrompt> --model <model> --workspace <path> --output-format text --trust --force
-```
-
-The workspace path is taken from `config.workspace` or extracted from the thread context via `runtime.extract`.
-
-## Hermes Agent
-
-`createHermesAgent(config)` invokes `hermes chat`:
-
-```
-hermes chat -q <fullPrompt> --yolo --max-turns 90 --quiet [--model <model>]
-```
-
-## LLM Adapter
-
-`createLlmAdapter(provider)` calls the OpenAI-compatible chat completions endpoint directly. It builds a two-message conversation (system + user) from the role's `systemPrompt` and `buildThreadInput` output, then extracts structured output from the response.
-
-## React Adapter
-
-`createReactAdapter(config)` creates a ReAct loop agent with four default tools: `read_file`, `write_file`, `patch_file`, and `shell_exec`. The loop continues until the agent calls the structured extraction tool or until `maxRounds` is exceeded.
-
-## Code Pointers
-
-| Package | File | What it does |
-|---------|------|-------------|
-| `@uncaged/workflow-protocol` | `src/types.ts` | `AdapterFn`, `AdapterBinding`, `RoleFn`, `RoleResult`, `AgentFn` |
-| `@uncaged/workflow-runtime` | `src/create-workflow.ts` | `createWorkflow` — dispatches `adapterForRole` each iteration |
-| `@uncaged/workflow-util-agent` | `src/build-agent-prompt.ts` | `buildThreadInput`, `buildAgentPrompt` |
-| `@uncaged/workflow-util-agent` | `src/spawn-cli.ts` | `spawnCli` — subprocess runner with timeout |
-| `@uncaged/workflow-util-agent` | `src/create-agent-adapter.ts` | `createAgentAdapter` — wraps `AgentFn` into `AdapterFn` |
-| `@uncaged/workflow-agent-cursor` | `src/index.ts` | `createCursorAgent` |
-| `@uncaged/workflow-agent-hermes` | `src/index.ts` | `createHermesAgent` |
-| `@uncaged/workflow-agent-llm` | `src/create-llm-adapter.ts` | `createLlmAdapter` |
-| `@uncaged/workflow-agent-react` | `src/create-react-adapter.ts` | `createReactAdapter` |
-
-## See Also
-
- [Role](./role.md) — the pure data that the binding executes
- [Engine](./engine.md) — the loop that invokes the bound adapter each step
@@ -1,83 +0,0 @@
-# Bundle
-
-> A self-contained single-file ESM module (`.esm.js`) that implements one workflow, identified by its XXH64 hash encoded as 13-char Crockford Base32.
-
-## Overview
-
-A bundle is the physical unit of workflow distribution. Workflow authors build their TypeScript source into a single ESM file using `bun build` with `@uncaged/*` packages as externals. The resulting `.esm.js` is the artifact that gets registered and executed.
-
-Every bundle is immutable and content-addressed: its identity is the XXH64 hash of its bytes, encoded as 13 characters of Crockford Base32 (e.g., `3TNKQRJ7BM4XH`). Registering a bundle with a new version simply adds a new hash entry; old hashes stay in the registry history and remain valid.
-
-Bundles are stored on disk at `~/.uncaged/workflow/bundles/<hash>/` after registration. The `cas/` and `threads.json` for that bundle's execution state live under the same directory.
-
-## Exports
-
-Every valid bundle must export exactly two named exports — no default export is permitted:
-
-| Export | Type | Description |
-|--------|------|-------------|
-| `run` | `WorkflowFn` | The `AsyncGenerator` that drives the execution loop |
-| `descriptor` | `WorkflowDescriptor` | Serializable metadata: description, roles, and routing graph |
-
-```typescript
-// Minimal bundle shape
-export const run: WorkflowFn = createWorkflow(def, binding);
-export const descriptor: WorkflowDescriptor = buildDescriptor(def);
-```
-
-The validator in `@uncaged/workflow-register` enforces this contract before a bundle can be registered — see `extractBundleExports`.
-
-## Hash Algorithm
-
-The bundle hash is computed with **XXH64** (seed 0) over the raw bytes of the `.esm.js` file, then encoded as 13-char Crockford Base32 using `encodeUint64AsCrockford`:
-
-```typescript
-// packages/workflow-cas/src/hash.ts
-export function hashWorkflowBundleBytes(data: Uint8Array): string {
-  const buf = Buffer.from(data.buffer, data.byteOffset, data.byteLength);
-  const digest = XXH.h64(0).update(buf).digest();
-  return encodeUint64AsCrockford(digestToUint64(digest));
-}
-```
-
-The same algorithm hashes CAS blob content (`hashString`), so all IDs in the system are consistent Crockford Base32 strings.
-
-## Build Process
-
-Bundles are not distributed from the monorepo directly. The typical flow is:
-
-1. Create a separate workspace (e.g., `my-workflows/`) with `@uncaged/workflow-runtime` as a dependency.
-2. Write a TypeScript workflow module that imports `createWorkflow` from `@uncaged/workflow-runtime`.
-3. Run `bun build --entrypoints src/my-workflow.ts --outfile dist/my-workflow.esm.js --format esm --external '@uncaged/*'`.
-4. Register with `uncaged-workflow workflow add <name> dist/my-workflow.esm.js`.
-
-## Storage Layout
-
-```
-~/.uncaged/workflow/
-  workflow.yaml                     # registry (name → hash mapping)
-  bundles/
-    <hash>/
-      threads.json                  # active thread index
-      history/
-        YYYY-MM-DD.jsonl            # completed thread records
-  cas/
-    <hash>.txt                      # CAS blobs (all bundles share one global CAS)
-```
-
-## Code Pointers
-
-| Package | File | What it does |
-|---------|------|-------------|
-| `@uncaged/workflow-cas` | `src/hash.ts` | `hashWorkflowBundleBytes` and `hashString` — XXH64 + Crockford encoding |
-| `@uncaged/workflow-register` | `src/bundle/extract-bundle-exports.ts` | Loads a `.esm.js` bundle and validates `run` + `descriptor` |
-| `@uncaged/workflow-register` | `src/bundle/bundle-validator.ts` | Schema validation of bundle exports |
-| `@uncaged/workflow-runtime` | `src/create-workflow.ts` | `createWorkflow` — the primary bundle authoring function |
-| `@uncaged/workflow-util` | `src/base32.ts` | `encodeUint64AsCrockford` — Crockford Base32 encoding |
-| `@uncaged/workflow-util` | `src/storage-root.ts` | `getDefaultWorkflowStorageRoot` → `~/.uncaged/workflow` |
-
-## See Also
-
- [Registry](./registry.md) — how bundles are registered and named in `workflow.yaml`
- [Thread](./thread.md) — how a bundle's `run` export is executed as a thread
- [Engine](./engine.md) — the executor that drives the bundle's `AsyncGenerator`
@@ -1,111 +0,0 @@
-# CAS (Content-Addressable Storage)
-
-> An append-only store where every blob is identified by its XXH64 hash, used to persist all workflow thread state as immutable Merkle nodes.
-
-## Overview
-
-CAS is the persistence substrate for the entire workflow engine. Rather than mutating a database row, every piece of state — agent output, role metadata, thread start parameters — is serialized as a YAML blob and stored under its hash. Because content determines identity, the same content always maps to the same hash, and writes are idempotent.
-
-The `CasStore` interface is intentionally simple: `put`, `get`, `delete`, `list`. The default filesystem implementation stores each blob as `<hash>.txt` under `~/.uncaged/workflow/cas/`. Writes use an atomic rename-from-tmp pattern to prevent partial writes.
-
-## Hash Algorithm
-
-All hashes in the system are **XXH64** (seed 0) over UTF-8 content, encoded as 13-char Crockford Base32. This applies to both CAS blob hashes and bundle file hashes. The encoding function `encodeUint64AsCrockford` lives in `@uncaged/workflow-util`.
-
-## Node Types
-
-The CAS holds three types of YAML nodes, all sharing the `{ type, payload, refs }` envelope:
-
-### `content` node
-Stores the raw text output of an agent or the initial prompt. `refs` lists any artifact hashes the content references.
-
-```yaml
-type: content
-payload: "The implementation is complete. Changed files: src/foo.ts"
-refs:
-  - 3TNKQRJ7BM4XH   # optional artifact refs
-```
-
-### `start` node
-Written once when a thread begins. Anchors the thread to a specific workflow name, bundle hash, and depth level.
-
-```yaml
-type: start
-payload:
-  name: solve-issue
-  hash: 3TNKQRJ7BM4XH
-  depth: 0
-  parentState: null
-refs:
-  - <promptHash>
-```
-
-### `state` node
-Written once per completed role step. Points back to the `start` node, the role's content node, and maintains an ancestor skip-list for traversal.
-
-```yaml
-type: state
-payload:
-  role: coder
-  meta: { status: "done", completedPhase: "..." }
-  start: <startHash>
-  content: <contentHash>
-  ancestors: [<prev_state>, ...]
-  compact: null
-  timestamp: 1716000000000
-  childThread: null
-refs:
-  - <contentHash>
-  - <startHash>
-  - <ancestor hashes>
-```
-
-## Merkle Structure
-
-The `ancestors` array in each `StateNode` implements a **skip-list** capped at 11 entries (1 direct parent + up to 10 skip-list ancestors). This allows `O(log n)` traversal of the chain without loading every node, while keeping each blob self-contained.
-
-```mermaid
-graph LR
-    S[StartNode] --> C1[content₁]
-    N1[StateNode₁] --> S
-    N1 --> C1
-    N2[StateNode₂] --> N1
-    N2 --> S
-    N2 --> C2[content₂]
-    END[StateNode __end__] --> N2
-    END --> S
-```
-
-## CasStore Interface
-
-```typescript
-type CasStore = {
-  put(content: string): Promise<string>;    // returns hash
-  get(hash: string): Promise<string | null>;
-  delete(hash: string): Promise<void>;
-  list(): Promise<string[]>;
-};
-```
-
-`put` normalizes raw strings into `content` Merkle nodes before hashing; pre-serialized RFC v3 nodes pass through unchanged.
-
-## Garbage Collection
-
-`cas gc` performs a mark-and-sweep over all CAS blobs. It seeds the reachable set from `head` and `start` hashes in every `threads.json` and `history/*.jsonl`, then traverses `refs` edges transitively. Unreachable blobs are deleted. The result reports `scannedThreads`, `activeRefs`, and `deletedEntries`.
-
-## Code Pointers
-
-| Package | File | What it does |
-|---------|------|-------------|
-| `@uncaged/workflow-protocol` | `src/types.ts` | `CasStore` interface definition |
-| `@uncaged/workflow-protocol` | `src/cas-types.ts` | `StartNode`, `StateNode`, `ContentMerkleNode` types |
-| `@uncaged/workflow-cas` | `src/cas.ts` | `createCasStore` — filesystem implementation |
-| `@uncaged/workflow-cas` | `src/hash.ts` | `hashString`, `hashWorkflowBundleBytes` — XXH64 + Crockford |
-| `@uncaged/workflow-cas` | `src/nodes.ts` | `putStartNode`, `putStateNode`, `putContentNodeWithRefs`, `parseCasThreadNode` |
-| `@uncaged/workflow-cas` | `src/merkle.ts` | `parseMerkleNode`, `serializeMerkleNode`, `getContentMerklePayload` |
-| `@uncaged/workflow-cas` | `src/reachable.ts` | Reachability traversal for GC |
-| `@uncaged/workflow-execute` | `src/engine/gc.ts` | GC orchestration |
-
-## See Also
-
- [Thread](./thread.md) — how thread execution state maps to CAS nodes
@@ -1,107 +0,0 @@
-# CLI
-
-> `uncaged-workflow` — the command-line tool for registering bundles, running threads, inspecting CAS, and connecting to the gateway.
-
-## Overview
-
-The CLI (`@uncaged/cli-workflow`) is the primary human interface to the workflow engine. It is a multi-level command dispatcher: top-level command groups (`workflow`, `thread`, `cas`, `init`, `setup`) each have a set of subcommands. Two shortcuts (`run`, `live`) alias frequently-used subcommands.
-
-The storage root defaults to `~/.uncaged/workflow` and can be overridden with `WORKFLOW_STORAGE_ROOT` or `UNCAGED_WORKFLOW_STORAGE_ROOT` environment variables.
-
-## Command Reference
-
-### Workflow Registry (`workflow`)
-
-| Subcommand | Args | Description |
-|-----------|------|-------------|
-| `workflow add` | `<name> <file.esm.js> [--types <path>]` | Register a workflow bundle in the registry |
-| `workflow list` | | List all registered workflows |
-| `workflow show` | `<name>` | Show bundle hash, timestamp, and descriptor |
-| `workflow rm` | `<name>` | Remove a workflow from the registry |
-| `workflow history` | `<name>` | Show version history for a workflow |
-| `workflow rollback` | `<name> [hash]` | Roll back to a previous version |
-
-### Thread Execution (`thread`)
-
-| Subcommand | Args | Description |
-|-----------|------|-------------|
-| `thread run` | `<name> [--prompt <text>]` | Start a new thread for a workflow; prints thread ID |
-| `thread list` | `[name]` | List threads, optionally filtered by workflow name |
-| `thread show` | `<id>` | Show thread steps and state from CAS |
-| `thread rm` | `<id>` | Remove a thread (from index and history) |
-| `thread fork` | `<thread-id> [--from-role <role>]` | Fork from an existing thread |
-| `thread ps` | | List running (active) threads |
-| `thread kill` | `<thread-id>` | Send kill signal to a running thread |
-| `thread live` | `<thread-id> \| --latest [--debug] [--role <name>]` | Attach and stream output live |
-| `thread pause` | `<thread-id>` | Pause a running thread |
-| `thread resume` | `<thread-id>` | Resume a paused thread |
-
-### CAS Inspection (`cas`)
-
-| Subcommand | Args | Description |
-|-----------|------|-------------|
-| `cas get` | `<hash>` | Print a CAS blob by hash |
-| `cas put` | `<content>` | Store content in CAS, print hash |
-| `cas list` | | List all hashes in CAS |
-| `cas rm` | `<hash>` | Remove a CAS entry |
-| `cas gc` | | Garbage-collect unreferenced entries |
-
-### Other Commands
-
-| Command | Args | Description |
-|---------|------|-------------|
-| `run <name> [...]` | | Shortcut for `thread run` |
-| `live <id> [...]` | | Shortcut for `thread live` |
-| `init` | | Scaffold a workflow workspace |
-| `setup` | | Configure LLM providers in `workflow.yaml` |
-| `connect [--name NAME] [--gateway URL]` | | Connect to gateway via WebSocket |
-| `skill [topic]` | | Print agent-consumable docs (`cli`, `develop`, `author`) |
-
-## Common Usage Examples
-
-```bash
-# Register a bundle
-uncaged-workflow workflow add solve-issue dist/solve-issue.esm.js
-
-# Run a workflow (prints thread ID)
-uncaged-workflow run solve-issue --prompt "Fix the login bug in auth.ts"
-
-# Watch live output
-uncaged-workflow live <thread-id>
-
-# Inspect a CAS blob
-uncaged-workflow cas get 3TNKQRJ7BM4XH
-
-# Show all running threads
-uncaged-workflow thread ps
-
-# Garbage-collect
-uncaged-workflow cas gc
-
-# Roll back to previous version
-uncaged-workflow workflow rollback solve-issue
-```
-
-## Environment Variables
-
-| Variable | Description |
-|----------|-------------|
-| `WORKFLOW_STORAGE_ROOT` | Override storage directory (default: `~/.uncaged/workflow`) |
-| `UNCAGED_WORKFLOW_STORAGE_ROOT` | Internal override; takes priority over `WORKFLOW_STORAGE_ROOT` |
-
-## Code Pointers
-
-| Package | File | What it does |
-|---------|------|-------------|
-| `@uncaged/cli-workflow` | `src/cli-dispatch.ts` | Top-level command router (`COMMAND_TABLE`) |
-| `@uncaged/cli-workflow` | `src/cli-usage.ts` | Usage text formatting |
-| `@uncaged/cli-workflow` | `src/commands/workflow/dispatch.ts` | `WORKFLOW_SUBCOMMAND_TABLE` |
-| `@uncaged/cli-workflow` | `src/commands/thread/dispatch.ts` | `THREAD_SUBCOMMAND_TABLE` |
-| `@uncaged/cli-workflow` | `src/commands/cas/dispatch.ts` | `CAS_SUBCOMMAND_TABLE` |
-| `@uncaged/cli-workflow` | `src/cli.ts` | CLI entry point |
-
-## See Also
-
- [Bundle](./bundle.md) — what `workflow add` registers
- [Thread](./thread.md) — what `thread run` creates
- [Registry](./registry.md) — the `workflow.yaml` that `workflow` commands manage
@@ -1,74 +0,0 @@
-# Dashboard
-
-> A private React single-page application for browsing workflows, inspecting thread execution records, and triggering runs via a connected gateway.
-
-## Overview
-
-The dashboard (`workflow-dashboard`) is a read-mostly web UI that surfaces thread history and workflow metadata. It is a private package (not published to npm) and is deployed separately from the CLI. It communicates with one or more remote workflow engine instances through the `workflow-gateway` WebSocket gateway, which proxies API calls back to each connected CLI client.
-
-The dashboard is not required to use the workflow engine — it is an optional observability layer on top of the same data that the CLI exposes.
-
-## Tech Stack
-
-| Concern | Choice |
-|---------|--------|
-| Framework | React (functional components, hooks) |
-| Build | Vite |
-| Styling | CSS variables via Tailwind-compatible utility classes |
-| Charts/graphs | ReactFlow (workflow graph visualization) |
-| HTTP | Native `fetch` with Bearer token auth |
-| Transport | REST over HTTP (proxied through the gateway) |
-
-## Data Sources
-
-The dashboard consumes four REST endpoints per connected client (proxied by the gateway):
-
-| Endpoint | Data |
-|----------|------|
-| `GET /workflows` | List of registered workflows with current hash and timestamp |
-| `GET /workflows/:name` | Full workflow detail including `WorkflowDescriptor` and version history |
-| `GET /threads` | All threads (active + completed) with summary fields |
-| `GET /threads/:id` | Thread records: `ThreadStartRecord`, `RoleRecord[]`, `WorkflowResultRecord` |
-
-The gateway multiplexes multiple CLI clients; the sidebar allows switching between them.
-
-## Views
-
-| View | Description |
-|------|-------------|
-| **Workflows** | Lists all registered workflows; clicking shows hash, descriptor, role graph, and version history |
-| **Threads** | Lists all threads; clicking shows the full step-by-step execution record with role metadata |
-| **Run dialog** | Form to start a new thread by picking a workflow and entering a prompt |
-
-### Workflow Graph
-
-Each workflow's `WorkflowDescriptor.graph` is rendered as an interactive ReactFlow diagram. Nodes represent roles (plus `__start__` and `__end__` terminals); edges represent moderator transitions labeled with condition names.
-
-## Authentication
-
-A Bearer token (stored in `localStorage` under `workflow-api-key`) is sent with every API request. The login page prompts for this key on first load. The gateway validates the token before proxying requests to connected clients.
-
-## Gateway Connection
-
-`uncaged-workflow connect [--name NAME] [--gateway URL]` registers the local workflow engine as a named client with the gateway over a WebSocket. The gateway then forwards REST API calls from the dashboard to the connected CLI process. The dashboard calls `GET /api/gateway/endpoints` to discover connected clients.
-
-## Private App Status
-
-`workflow-dashboard` has `"private": true` in its `package.json` and is excluded from the changeset versioning pipeline. It is developed alongside the engine packages but distributed separately (e.g., as a static build hosted alongside the gateway server).
-
-## Code Pointers
-
-| Package | File | What it does |
-|---------|------|-------------|
-| `workflow-dashboard` | `src/app.tsx` | Root component — routing, auth state, view switching |
-| `workflow-dashboard` | `src/api.ts` | All API functions + endpoint types (`ThreadRecord`, `WorkflowDetail`, etc.) |
-| `workflow-dashboard` | `src/components/thread-detail.tsx` | Thread step viewer |
-| `workflow-dashboard` | `src/components/workflow-graph/workflow-graph.tsx` | ReactFlow graph of workflow roles and transitions |
-| `workflow-dashboard` | `src/components/sidebar.tsx` | Client selector and view navigation |
-| `@uncaged/workflow-gateway` | `src/index.ts` | Gateway server entry point |
-| `@uncaged/workflow-gateway` | `src/ws-protocol.ts` | WebSocket message protocol between CLI and gateway |
-
-## See Also
-
- [Thread](./thread.md) — the execution records the dashboard displays
- [Engine](./engine.md) — the process that produces those records
@@ -1,110 +0,0 @@
-# Engine
-
-> The execution loop that drives a workflow bundle's `AsyncGenerator`, persisting each yielded `RoleOutput` as a CAS `StateNode` and managing thread lifecycle.
-
-## Overview
-
-The engine (`executeThread`) takes a `WorkflowFn` and runs it to completion. It is responsible for three concerns: persisting each role output to CAS, updating the active-thread index after every step, and terminating the thread cleanly when the generator finishes, is aborted, or is killed by the supervisor.
-
-The engine does not interact with LLMs directly — that responsibility belongs to the workflow bundle's `run` function and its bound agent adapters. The engine only observes `RoleOutput` values yielded by the generator.
-
-## Execution Flow
-
-```mermaid
-flowchart TD
-    A[executeThread] --> B[putStartNode → CAS]
-    B --> C[publishHead → threads.json]
-    C --> D{generator.next}
-    D -- done --> E[finalizeThread]
-    D -- yield RoleOutput --> F[appendStateForStep → CAS]
-    F --> G[publishHead → threads.json]
-    G --> H{supervisorInterval?}
-    H -- kill --> E
-    H -- continue --> I{awaitAfterEachYield}
-    I --> D
-    D -- AbortSignal --> J[finalizeAbortedThread]
-    E --> K[removeThreadEntry]
-    K --> L[appendThreadHistoryEntry]
-```
-
-## Role Loop (inside the bundle's `createWorkflow`)
-
-The `WorkflowFn` produced by `createWorkflow` runs its own loop — one iteration per role step:
-
-1. **Moderator**: calls `pickNext(ctx)` (derived from the `ModeratorTable`) → returns a role name or `END`.
-2. **Adapter**: calls the bound `AdapterFn` with the role's `systemPrompt` and Zod schema → returns `RoleFn` → executes → returns `RoleResult<T>`.
-3. **Persist**: calls `putContentNodeWithRefs` to store the role output in CAS, constructs a `RoleStep`, and `yield`s a `RoleOutput` to the engine.
-
-```mermaid
-sequenceDiagram
-    participant E as Engine
-    participant W as WorkflowFn (bundle)
-    participant M as Moderator
-    participant A as AdapterFn
-    participant C as CAS
-
-    E->>W: generator.next()
-    W->>M: pickNext(ctx) → roleName
-    W->>A: adapter(systemPrompt, schema)(ctx, runtime)
-    A-->>W: RoleResult { meta, childThread }
-    W->>C: putContentNodeWithRefs(JSON.stringify(meta))
-    W-->>E: yield RoleOutput
-    E->>C: putStateNode(StateNodePayload)
-    E->>E: publishHead(threads.json)
-```
-
-## Key Types
-
-```typescript
-// Engine input
-type ExecuteThreadOptions = {
-  depth: number;
-  parentStateHash: string | null;
-  signal: AbortSignal;
-  awaitAfterEachYield: () => Promise<void>;  // used for pause/resume gate
-  forkContinuation: ForkContinuationOptions | null;
-  prefilledDiskSteps: PrefilledDiskStep[] | null;
-  replayTimestamps: readonly number[] | null;
-  storageRoot: string;
-};
-
-// Engine output
-type WorkflowResult = {
-  returnCode: number;
-  summary: string;
-  rootHash: string;   // hash of the __end__ StateNode
-};
-```
-
-## Pause Gate
-
-`awaitAfterEachYield` is a function injected by the worker/runner that can block the loop between steps. The `ThreadPauseGate` in `thread-pause-gate.ts` provides `pause()` / `resume()` operations that control this gate. When paused, the loop suspends after writing the current step but before requesting the next one.
-
-## Supervisor
-
-If `workflowConfig.supervisorInterval > 0`, the engine runs a supervisor check after every `supervisorInterval` steps. The supervisor calls an LLM with a summary of recent steps and returns `"continue"` or `"kill"`. A `"kill"` decision finalizes the thread immediately with `returnCode: 1` and a summary string.
-
-## Summarizer
-
-On normal completion (generator returns), the engine calls `createSummarizer` to produce a single LLM-generated summary string from recent step content. This summary replaces the bundle's raw `WorkflowCompletion.summary` in the final history record.
-
-## Code Pointers
-
-| Package | File | What it does |
-|---------|------|-------------|
-| `@uncaged/workflow-execute` | `src/engine/engine.ts` | `executeThread` — main engine entry point |
-| `@uncaged/workflow-execute` | `src/engine/types.ts` | `ExecuteThreadOptions`, `ExecuteThreadIo`, `ChainState`, `ThreadPauseGate` |
-| `@uncaged/workflow-execute` | `src/engine/threads-index.ts` | `threads.json` persistence, history append |
-| `@uncaged/workflow-execute` | `src/engine/supervisor.ts` | Supervisor LLM check (`"continue"` / `"kill"`) |
-| `@uncaged/workflow-execute` | `src/engine/summarizer.ts` | Post-completion LLM summary |
-| `@uncaged/workflow-execute` | `src/engine/thread-pause-gate.ts` | Pause/resume gate |
-| `@uncaged/workflow-execute` | `src/engine/worker.ts` | Worker-process entry that spawns `executeThread` in a subprocess |
-| `@uncaged/workflow-runtime` | `src/create-workflow.ts` | `createWorkflow` — the role loop inside the bundle |
-| `@uncaged/workflow-protocol` | `src/types.ts` | `WorkflowFn`, `RoleOutput`, `WorkflowCompletion`, `AdvanceOutcome` |
-
-## See Also
-
- [Role](./role.md) — what the moderator selects each iteration
- [Agent Binding](./agent-binding.md) — what executes a role and returns its output
- [Reactor](./reactor.md) — used internally for the extract and supervisor LLM calls
- [Thread](./thread.md) — the CAS-persisted result of running the engine
@@ -1,129 +0,0 @@
-# Package Map
-
-> All packages in the monorepo with their responsibilities, dependency layers, and publication status.
-
-## Overview
-
-The monorepo is organized as a strict dependency DAG. Each layer may only depend on layers below it. The execution stack flows from the shared protocol types at the bottom up to the CLI at the top. Agent packages and template packages are leaf nodes that depend on the runtime layer but are not depended upon by the core stack.
-
-## Package List
-
-| Package | Description |
-|---------|-------------|
-| `@uncaged/workflow-protocol` | Shared types (`ThreadContext`, `RoleDefinition`, `CasStore`, `Result`, etc.) and constants (`START`, `END`) |
-| `@uncaged/workflow-runtime` | `createWorkflow`, type re-exports; primary dependency for bundle authors |
-| `@uncaged/workflow-util` | Utilities: Crockford Base32, ULID, structured logger, storage paths |
-| `@uncaged/workflow-reactor` | `createThreadReactor` (ReAct loop), `createLlmFn` (OpenAI-compatible LLM caller) |
-| `@uncaged/workflow-cas` | `createCasStore` (filesystem CAS), XXH64 hashing, Merkle node serialization |
-| `@uncaged/workflow-register` | Bundle validation, `workflow.yaml` registry read/write, model resolution |
-| `@uncaged/workflow-execute` | Engine (`executeThread`), extract phase, fork, GC, `workflowAsAgent` |
-| `@uncaged/cli-workflow` | `uncaged-workflow` CLI — command dispatcher for all user-facing operations |
-| `@uncaged/workflow-agent-cursor` | Adapter that runs the `cursor` CLI non-interactively in a workspace |
-| `@uncaged/workflow-agent-hermes` | Adapter that runs `hermes chat` (Nerve-style CLI agent) |
-| `@uncaged/workflow-agent-llm` | Adapter for direct LLM chat completions |
-| `@uncaged/workflow-agent-react` | Adapter with ReAct loop and file/shell tools |
-| `@uncaged/workflow-util-agent` | Shared agent utilities: `buildThreadInput`, `spawnCli`, `createAgentAdapter` |
-| `@uncaged/workflow-template-develop` | `develop` workflow template (planner → coder → reviewer → tester → committer) |
-| `@uncaged/workflow-template-solve-issue` | `solve-issue` workflow template (preparer → developer → submitter) |
-| `@uncaged/workflow-gateway` | WebSocket gateway for remote CLI-to-dashboard communication |
-| `workflow-dashboard` | React dashboard (private, unpublished) — thread/workflow viewer |
-
-## Dependency Layer Diagram
-
-```mermaid
-graph TD
-    subgraph Layer 0 — Protocol
-        P[workflow-protocol]
-    end
-
-    subgraph Layer 1 — Foundations
-        RT[workflow-runtime]
-        UT[workflow-util]
-        RX[workflow-reactor]
-    end
-
-    subgraph Layer 2 — Storage & Register
-        CAS[workflow-cas]
-        REG[workflow-register]
-    end
-
-    subgraph Layer 3 — Execute
-        EX[workflow-execute]
-    end
-
-    subgraph Layer 4 — CLI
-        CLI[cli-workflow]
-    end
-
-    subgraph Agents (leaf)
-        AGC[workflow-agent-cursor]
-        AGH[workflow-agent-hermes]
-        AGL[workflow-agent-llm]
-        AGR[workflow-agent-react]
-        UA[workflow-util-agent]
-    end
-
-    subgraph Templates (leaf)
-        TD[workflow-template-develop]
-        TS[workflow-template-solve-issue]
-    end
-
-    subgraph Dashboard
-        GW[workflow-gateway]
-        DB[workflow-dashboard]
-    end
-
-    RT --> P
-    UT --> P
-    RX --> P
-    CAS --> P
-    REG --> P
-    REG --> UT
-    EX --> RT
-    EX --> UT
-    EX --> CAS
-    EX --> REG
-    EX --> RX
-    CLI --> EX
-    CLI --> UT
-    CLI --> REG
-    AGC --> RT
-    AGC --> UT
-    AGC --> UA
-    AGH --> RT
-    AGH --> UA
-    AGL --> RT
-    AGR --> RT
-    AGR --> RX
-    UA --> RT
-    TD --> RT
-    TS --> RT
-    DB --> GW
-```
-
-## Published vs. Private
-
-All `@uncaged/*` packages are published to **npmjs.org** under a fixed versioning scheme (all packages share the same version number via `@changesets/cli` in fixed mode).
-
-| Status | Packages |
-|--------|---------|
-| **Published** | All packages with `@uncaged/` scope |
-| **Private** | `workflow-dashboard` (no `@uncaged/` scope, `"private": true`) |
-
-## Code Pointers
-
-| Package | File | What it does |
-|---------|------|-------------|
-| `@uncaged/workflow-protocol` | `src/types.ts` | Root type definitions for the entire stack |
-| `@uncaged/workflow-runtime` | `src/index.ts` | Public API for bundle authors |
-| `@uncaged/workflow-util` | `src/index.ts` | Utility re-exports |
-| `@uncaged/workflow-execute` | `src/index.ts` | Engine public API |
-| `@uncaged/cli-workflow` | `src/cli-dispatch.ts` | Top-level command table |
-
-## See Also
-
- [Bundle](./bundle.md) — produced by workspace authors using `@uncaged/workflow-runtime`
- [Engine](./engine.md) — the core of `@uncaged/workflow-execute`
- [Reactor](./reactor.md) — `@uncaged/workflow-reactor`
- [Registry](./registry.md) — `@uncaged/workflow-register`
- [CLI](./cli.md) — `@uncaged/cli-workflow`
@@ -1,102 +0,0 @@
-# Reactor
-
-> A generic ReAct (Reason + Act) loop that drives an LLM through multiple tool-call rounds until it produces structured output matching a Zod schema.
-
-## Overview
-
-The reactor is a reusable abstraction for LLM interactions that require tool use. It runs a multi-turn conversation loop: the LLM is presented with a user message and a set of tools, and responds either with a tool call (which the reactor dispatches and feeds back) or with a plain JSON object matching the expected schema. The loop repeats until structured output is obtained or `maxRounds` is exhausted.
-
-The reactor is used in two places:
-
-1. **Extract phase** — `createExtract` in `@uncaged/workflow-execute` uses a CAS-backed reactor to extract typed `meta` from a role's content hash.
-2. **React agent** — `createReactAdapter` in `@uncaged/workflow-agent-react` uses the reactor as its execution backbone.
-
-## createThreadReactor
-
-```typescript
-function createThreadReactor<TThread>(
-  config: ThreadReactorConfig<TThread>,
-): ThreadReactorFn<TThread>
-```
-
-`ThreadReactorConfig` bundles:
-
-| Field | Purpose |
-|-------|---------|
-| `llm` | The `LlmFn` to call each round |
-| `staticTools` | Tools always available (e.g., `cas_get`) |
-| `structuredToolFromSchema` | Derives a schema-specific extraction tool from the Zod schema |
-| `systemPromptForStructuredTool` | Constructs the system prompt given the extraction tool name |
-| `toolHandler` | Handles non-structured tool calls; receives the raw `ToolCall` and thread context |
-| `maxRounds` | Hard stop after N rounds; returns `err("max_react_rounds_exceeded")` |
-
-## Round Lifecycle
-
-```mermaid
-sequenceDiagram
-    participant R as Reactor
-    participant L as LLM
-    participant H as toolHandler
-
-    R->>L: messages + tools
-    L-->>R: response
-
-    alt plain JSON (valid schema)
-        R-->>R: return ok(value)
-    else plain JSON (invalid)
-        R->>L: correction message
-    else tool_calls
-        loop each call
-            alt structured tool
-                R-->>R: validate args → return ok(value)
-            else static tool
-                R->>H: toolHandler(call, thread)
-                H-->>R: content string
-                R->>L: tool result message
-            end
-        end
-    end
-```
-
-## LlmFn
-
-```typescript
-type LlmFn = (input: {
-  messages: ChatMessage[];
-  tools: readonly ToolDefinition[];
-}) => Promise<Result<string, string>>;
-```
-
-`createLlmFn(provider)` in `@uncaged/workflow-reactor` builds an `LlmFn` that calls the OpenAI-compatible chat completions endpoint and returns the raw response body as a string for the reactor to parse.
-
-## Extract Phase
-
-`createExtract(provider, { cas })` in `@uncaged/workflow-execute` creates a `CasReactor` — a preconfigured `ThreadReactorFn` with a `cas_get` static tool. The extract function loads the content payload for a given hash, sends it to the reactor with the role's Zod schema, and returns `ExtractResult<T>`.
-
-```typescript
-type ExtractFn = <T extends Record<string, unknown>>(
-  schema: z.ZodType<T>,
-  contentHash: string,
-) => Promise<ExtractResult<T>>;
-```
-
-The `cas_get` tool allows the LLM to dereference CAS hashes during extraction — important when the content node references artifact hashes.
-
-## Relationship to Engine
-
-The reactor is called within `AdapterFn` implementations (e.g., `createLlmAdapter`, `createReactAdapter`) when the agent needs multi-turn tool interaction to complete a role. The engine itself does not call the reactor directly — it only drives the outer `WorkflowFn` generator and persists `RoleOutput` values.
-
-## Code Pointers
-
-| Package | File | What it does |
-|---------|------|-------------|
-| `@uncaged/workflow-reactor` | `src/thread-reactor.ts` | `createThreadReactor` — generic ReAct loop |
-| `@uncaged/workflow-reactor` | `src/llm-fn.ts` | `createLlmFn` — OpenAI-compatible LLM caller |
-| `@uncaged/workflow-reactor` | `src/types.ts` | `LlmFn`, `ThreadReactorConfig`, `ToolCall`, `ToolDefinition`, `ChatMessage` |
-| `@uncaged/workflow-execute` | `src/cas-reactor.ts` | `createCasReactor` — reactor with `cas_get` static tool |
-| `@uncaged/workflow-execute` | `src/extract/extract-fn.ts` | `createExtract` — extract phase using the CAS reactor |
-
-## See Also
-
- [Engine](./engine.md) — drives the workflow generator; extract is called inside the adapter layer
- [Agent Binding](./agent-binding.md) — adapter implementations that use the reactor internally
@@ -1,95 +0,0 @@
-# Registry
-
-> `workflow.yaml` — the local file that maps workflow names to their current and historical bundle hashes, plus global LLM provider configuration.
-
-## Overview
-
-The registry is a single YAML file at `<storageRoot>/workflow.yaml` (default: `~/.uncaged/workflow/workflow.yaml`). It is the authoritative index of which bundles are available on a machine and what name each one is known by. All CLI workflow commands read or write this file.
-
-The registry is read on every `uncaged-workflow run` invocation to look up the bundle hash for a given name, then used again to resolve the `extract` model configuration. It is written atomically via the `writeWorkflowRegistry` function.
-
-## Schema
-
-```yaml
-config:
-  maxDepth: 3
-  supervisorInterval: 5
-  providers:
-    openrouter:
-      baseUrl: "https://openrouter.ai/api/v1"
-      apiKey: "sk-or-..."
-  models:
-    extract: "openrouter/anthropic/claude-sonnet-4-5"
-    supervisor: "openrouter/anthropic/claude-haiku-3-5"
-
-workflows:
-  solve-issue:
-    hash: "3TNKQRJ7BM4XH"
-    timestamp: 1716000000000
-    history:
-      - hash: "2BMJPQ6YAK3WG"
-        timestamp: 1715000000000
-  develop:
-    hash: "7VQWX8NRHK1ZT"
-    timestamp: 1716100000000
-    history: []
-```
-
-## Types
-
-```typescript
-type WorkflowRegistryFile = {
-  config: WorkflowConfig | null;
-  workflows: Record<string, WorkflowRegistryEntry>;
-};
-
-type WorkflowRegistryEntry = {
-  hash: string;       // current bundle hash (13-char Crockford Base32)
-  timestamp: number;  // Unix epoch ms when this version was registered
-  history: WorkflowHistoryEntry[];
-};
-
-type WorkflowHistoryEntry = {
-  hash: string;
-  timestamp: number;
-};
-```
-
-## Bundle Registration Flow
-
-1. `uncaged-workflow workflow add <name> <file.esm.js>` is called.
-2. The bundle bytes are hashed with XXH64 → 13-char Crockford Base32.
-3. The bundle file is copied into `<storageRoot>/bundles/<hash>/` (if not already present).
-4. `registerWorkflowVersion` prepends the current head to `history` and sets the new hash as head.
-5. The updated registry is written back to `workflow.yaml`.
-
-## Version History
-
-Every `workflow add` on an already-registered name pushes the previous hash into `history`. History is ordered most-recent-first. `workflow rollback <name> [hash]` swaps the specified history entry back to head (or defaults to `history[0]`).
-
-## Model Resolution
-
-The `config.models` section uses `provider/model` references (e.g., `"openrouter/anthropic/claude-sonnet-4-5"`). `resolveModel` splits the reference on the first `/`, looks up the provider in `config.providers`, and returns a `ResolvedModel` with `{ baseUrl, apiKey, model }`. This is used by the engine to configure the `extract` LLM.
-
-```typescript
-// packages/workflow-register/src/config/resolve-model.ts
-export function resolveModel(
-  config: WorkflowConfig,
-  modelKey: string,
-): Result<ResolvedModel, string>
-```
-
-## Code Pointers
-
-| Package | File | What it does |
-|---------|------|-------------|
-| `@uncaged/workflow-register` | `src/registry/registry.ts` | `readWorkflowRegistry`, `writeWorkflowRegistry`, `registerWorkflowVersion`, `rollbackWorkflowToHistoryHash` |
-| `@uncaged/workflow-register` | `src/registry/types.ts` | `WorkflowRegistryFile`, `WorkflowRegistryEntry`, `WorkflowHistoryEntry` |
-| `@uncaged/workflow-register` | `src/registry/registry-normalize.ts` | YAML normalization for the registry root |
-| `@uncaged/workflow-register` | `src/config/resolve-model.ts` | `resolveModel` — splits `provider/model` refs |
-| `@uncaged/workflow-register` | `src/bundle/extract-bundle-exports.ts` | Validates bundle exports before registration |
-| `@uncaged/workflow-protocol` | `src/types.ts` | `WorkflowConfig`, `ProviderConfig`, `ResolvedModel` |
-
-## See Also
-
- [Bundle](./bundle.md) — what is stored and indexed in the registry
@@ -1,72 +0,0 @@
-# Role
-
-> A named actor within a workflow defined entirely as pure data — a description, a system prompt, an extraction schema, and an optional refs extractor — with no embedded agent logic.
-
-## Overview
-
-A role is a `RoleDefinition<Meta>` value: a plain TypeScript object that describes what an actor in the workflow does and how its output should be structured. Roles are authored in the template or bundle source and passed to `createWorkflow` as part of the `WorkflowDefinition`. They never hold a reference to an agent implementation.
-
-This separation of concerns is deliberate. The same role definition can be executed by different agents (Cursor, Hermes, an LLM, a React loop) simply by changing the `AdapterBinding` passed to `createWorkflow`. Roles are also serialized into the `WorkflowDescriptor` for tooling like the dashboard.
-
-## RoleDefinition Type
-
-```typescript
-type RoleDefinition<Meta extends Record<string, unknown>> = {
-  description: string;
-  systemPrompt: string;
-  schema: z.ZodType<Meta>;
-  extractRefs: ((meta: Meta) => string[]) | null;
-};
-```
-
-| Field | Purpose |
-|-------|---------|
-| `description` | Human-readable summary for tooling and the `WorkflowDescriptor` |
-| `systemPrompt` | Passed to the adapter as the agent's persona/instruction for this role |
-| `schema` | Zod v4 schema that defines the structured output (`Meta`) of the role |
-| `extractRefs` | Optional function that extracts CAS hashes from `meta` to record as artifact refs |
-
-## Schema and Extraction
-
-Each role's `schema` is a Zod v4 type parameterized to the role's `Meta` type. When a role executes via an `AdapterFn`, the adapter is responsible for producing a value that satisfies this schema directly (the `AdapterFn` receives the schema and system prompt and returns a `RoleFn` that yields `RoleResult<T>`).
-
-If `extractRefs` is non-null, the engine calls it on the completed `meta` to collect additional CAS hashes that should appear in the `StateNode.refs` skip-list, enabling traversal of artifacts produced by the role.
-
-## WorkflowDefinition
-
-Roles are collected into a `WorkflowDefinition<M>` alongside the moderator table:
-
-```typescript
-type WorkflowDefinition<M extends RoleMeta> = {
-  description: string;
-  roles: { [K in keyof M & string]: RoleDefinition<M[K]> };
-  table: ModeratorTable<M>;
-};
-```
-
-`M` is the `RoleMeta` map that binds each role name to its concrete `Meta` type. This gives full TypeScript type safety across the moderator, adapter, and CAS storage layers.
-
-## WorkflowRoleDescriptor (Serialized)
-
-The `WorkflowDescriptor` (stored in the bundle's `descriptor` export) contains a `roles` map of `WorkflowRoleDescriptor` objects — a JSON-serializable projection of each `RoleDefinition`:
-
-```typescript
-type WorkflowRoleDescriptor = {
-  description: string;
-  systemPrompt: string;
-  schema: WorkflowRoleSchema;  // JSON-compatible schema shape
-};
-```
-
-## Code Pointers
-
-| Package | File | What it does |
-|---------|------|-------------|
-| `@uncaged/workflow-protocol` | `src/types.ts` | `RoleDefinition`, `WorkflowDefinition`, `RoleMeta`, `WorkflowRoleDescriptor`, `WorkflowDescriptor` |
-| `@uncaged/workflow-runtime` | `src/create-workflow.ts` | Consumes `WorkflowDefinition` roles in the adapter dispatch loop |
-| `@uncaged/workflow-register` | `src/bundle/build-descriptor.ts` | Serializes `RoleDefinition[]` to `WorkflowDescriptor` |
-
-## See Also
-
- [Engine](./engine.md) — the loop that selects and executes roles
- [Agent Binding](./agent-binding.md) — the runtime binding that executes a role via a concrete agent
@@ -1,97 +0,0 @@
-# Thread
-
-> A single execution instance of a workflow, identified by a ULID, whose state is stored as a linked chain of immutable CAS nodes.
-
-## Overview
-
-A thread is the runtime envelope around one call to a workflow's `run` function. It carries a unique ULID (26-char Crockford Base32) and tracks the full sequence of role steps that have executed. Because all state is written to CAS as immutable blobs, threads are append-only and fully auditable.
-
-Every thread belongs to a specific workflow bundle (identified by hash). The engine writes a `StartNode` when the thread begins and one `StateNode` per completed role step — including a final `__end__` state on completion or abort. Steps accumulate in `ThreadContext.steps` and are replayed into the context whenever a thread is resumed.
-
-## Lifecycle
-
-```mermaid
-stateDiagram-v2
-    [*] --> Active: thread run / fork
-    Active --> Active: role step yielded
-    Active --> Paused: pause signal
-    Paused --> Active: resume signal
-    Active --> Completed: generator returns WorkflowCompletion
-    Active --> Aborted: kill signal / AbortSignal
-    Completed --> [*]: entry in history/*.jsonl
-    Aborted --> [*]: entry in history/*.jsonl (returnCode=130)
-```
-
-## Identity
-
-Thread IDs are ULIDs: 26-char Crockford Base32 strings composed of a 10-char timestamp prefix and a 16-char random suffix. Generated by `generateUlid` from `@uncaged/workflow-util`.
-
-## State Storage
-
-Thread state is stored entirely in CAS as a linked list of nodes:
-
-```
-StartNode  (type: "start")
-  payload: { name, hash, depth, parentState }
-  refs: [promptHash, parentState?]
-
-StateNode  (type: "state")  ← one per role step
-  payload: { role, meta, start, content, ancestors[], compact, timestamp, childThread }
-  refs: [contentHash, startHash, ancestor hashes...]
-
-StateNode  (type: "state", role: "__end__")  ← final node
-  payload: { returnCode, summary }
-```
-
-The `ancestors` array implements a skip-list (capped at 11 entries: 1 direct parent + up to 10 ancestors) to allow efficient traversal without loading every node in the chain.
-
-## Index Files
-
-| File | Purpose |
-|------|---------|
-| `<bundleDir>/threads.json` | Active thread index — maps `threadId → { head, start, updatedAt }` |
-| `<bundleDir>/history/YYYY-MM-DD.jsonl` | Completed thread records — one JSON line per completed/aborted thread |
-| `<storageRoot>/cas/` | All CAS blobs shared across all bundles |
-
-A thread is "active" while it appears in `threads.json`. On completion, its entry is removed from `threads.json` and a record appended to the appropriate `history/*.jsonl` file.
-
-## ThreadContext
-
-The `ThreadContext` type is the read-only view passed into every role and moderator call:
-
-```typescript
-type ThreadContext<M extends RoleMeta = RoleMeta> = {
-  threadId: string;
-  depth: number;
-  bundleHash: string;
-  start: StartStep;
-  steps: RoleStep<M>[];
-};
-```
-
-`depth` tracks nesting for sub-workflow invocations (workflow-as-agent). `steps` grows by one entry after each successful role execution.
-
-## Fork
-
-A thread can be forked from any completed role step via `thread fork <id> [--from-role <role>]`. The fork reuses the original `StartNode` (same `startHash`) and replays CAS steps up to the fork point before resuming the generator. The forked thread gets a new ULID.
-
-## Debug Logs
-
-Each thread writes structured JSONL debug logs to `.info.jsonl` in the bundle directory. Each log line is `{ tag, content, timestamp }` where `tag` is an 8-char Crockford Base32 call-site identifier.
-
-## Code Pointers
-
-| Package | File | What it does |
-|---------|------|-------------|
-| `@uncaged/workflow-protocol` | `src/types.ts` | `ThreadContext`, `StartStep`, `RoleStep`, `RoleMeta` types |
-| `@uncaged/workflow-protocol` | `src/cas-types.ts` | `StartNode`, `StartNodePayload`, `StateNode`, `StateNodePayload` |
-| `@uncaged/workflow-execute` | `src/engine/threads-index.ts` | `threads.json` read/write, history append, `ThreadIndexEntry` |
-| `@uncaged/workflow-execute` | `src/engine/engine.ts` | `executeThread` — starts, drives, and finalizes a thread |
-| `@uncaged/workflow-execute` | `src/engine/fork-thread.ts` | Fork logic |
-| `@uncaged/workflow-util` | `src/ulid.ts` | `generateUlid` — ULID generation |
-
-## See Also
-
- [CAS](./cas.md) — the storage layer that holds all thread state nodes
- [Engine](./engine.md) — the execution loop that drives the thread
- [Bundle](./bundle.md) — the workflow being executed in this thread
@@ -1,153 +0,0 @@
-# Workflow Templates
-
-> Pre-built `WorkflowDefinition` objects exported from `@uncaged/workflow-template-*` packages that bundle authors can import, customize, or use directly.
-
-## Overview
-
-Templates are the reference implementations of common workflow patterns. They export a complete `WorkflowDefinition<M>` — typed roles with Zod schemas, and a `ModeratorTable` — ready to be passed to `createWorkflow`. A bundle author imports a template definition, supplies an `AdapterBinding`, calls `createWorkflow`, and exports the result as `run`.
-
-Templates are published as regular `@uncaged/*` npm packages. They are not bundles themselves; they are TypeScript libraries that become part of a bundle when the author's workspace is built.
-
-## solve-issue Template
-
-**Package**: `@uncaged/workflow-template-solve-issue`
-
-Resolves an issue end-to-end by preparing the repository, delegating implementation to a nested `develop` workflow, and opening a pull request.
-
-### Roles
-
-| Role | Description |
-|------|-------------|
-| `preparer` | Reads the issue, clones/checks out the repo, sets up the environment |
-| `developer` | Delegates to the `develop` workflow via `workflowAsAgent` (child thread) |
-| `submitter` | Opens a pull request with the completed changes |
-
-### Moderator Table
-
-```
-__start__ → preparer → developer → submitter → __end__
-```
-
-Linear routing — each role runs exactly once in sequence.
-
-### Meta Types
-
-```typescript
-type SolveIssueMeta = {
-  preparer: PreparerMeta;
-  developer: DeveloperMeta;
-  submitter: SubmitterMeta;
-};
-```
-
-## develop Template
-
-**Package**: `@uncaged/workflow-template-develop`
-
-Plans an implementation in phases, codes each phase incrementally, reviews, verifies with tests/build/lint, and commits.
-
-### Roles
-
-| Role | Description |
-|------|-------------|
-| `planner` | Produces an ordered list of implementation phases with hashes |
-| `coder` | Implements one phase; reports `completedPhase` hash in meta |
-| `reviewer` | Reviews the accumulated changes; approves or requests changes |
-| `tester` | Runs tests/lint/build; reports `passed` or `failed` |
-| `committer` | Creates the final git commit |
-
-### Moderator Table
-
-```
-__start__ → planner
-planner   → __end__      (if status == "aborted")
-planner   → coder        (fallback)
-coder     → reviewer     (if allPhasesComplete)
-coder     → coder        (fallback — repeat per phase)
-reviewer  → tester       (if status == "approved")
-reviewer  → coder        (fallback — request changes)
-tester    → committer    (if status == "passed")
-tester    → coder        (fallback — fix failures)
-committer → __end__
-```
-
-### Meta Types
-
-```typescript
-type DevelopMeta = {
-  planner: PlannerMeta;
-  coder: CoderMeta;
-  reviewer: ReviewerMeta;
-  tester: TesterMeta;
-  committer: CommitterMeta;
-};
-```
-
-## Writing a Custom Template
-
-A minimal custom workflow:
-
-```typescript
-import { createWorkflow, type WorkflowDefinition, END, START } from "@uncaged/workflow-runtime";
-import { z } from "zod/v4";
-import type { AdapterBinding } from "@uncaged/workflow-runtime";
-
-type MyMeta = {
-  analyst: { summary: string; confidence: number };
-  writer: { report: string };
-};
-
-const def: WorkflowDefinition<MyMeta> = {
-  description: "Analyse then write a report.",
-  roles: {
-    analyst: {
-      description: "Analyses the input and produces a structured summary.",
-      systemPrompt: "You are an expert analyst...",
-      schema: z.object({ summary: z.string(), confidence: z.number() }),
-      extractRefs: null,
-    },
-    writer: {
-      description: "Writes the final report.",
-      systemPrompt: "You are a technical writer...",
-      schema: z.object({ report: z.string() }),
-      extractRefs: null,
-    },
-  },
-  table: {
-    [START]: [{ condition: "FALLBACK", role: "analyst" }],
-    analyst: [{ condition: "FALLBACK", role: "writer" }],
-    writer: [{ condition: "FALLBACK", role: END }],
-  },
-};
-
-// In the bundle entry point:
-export const run = createWorkflow(def, binding);
-export const descriptor = buildDescriptor(def);
-```
-
-## Template → Bundle Relationship
-
-Templates are TypeScript library packages, not bundles. To use a template:
-
-1. Install the template package from npm: `bun add @uncaged/workflow-template-develop`.
-2. Import the definition: `import { developWorkflowDefinition } from "@uncaged/workflow-template-develop"`.
-3. Supply an `AdapterBinding` and call `createWorkflow`.
-4. Build with `bun build` to produce `.esm.js`.
-5. Register with `uncaged-workflow workflow add`.
-
-## Code Pointers
-
-| Package | File | What it does |
-|---------|------|-------------|
-| `@uncaged/workflow-template-solve-issue` | `src/index.ts` | `solveIssueWorkflowDefinition`, role and moderator exports |
-| `@uncaged/workflow-template-solve-issue` | `src/roles.ts` | `SolveIssueMeta`, `solveIssueRoles` |
-| `@uncaged/workflow-template-solve-issue` | `src/moderator.ts` | `solveIssueTable` — linear transition table |
-| `@uncaged/workflow-template-develop` | `src/index.ts` | `developWorkflowDefinition`, role and moderator exports |
-| `@uncaged/workflow-template-develop` | `src/roles.ts` | `DevelopMeta`, `developRoles` |
-| `@uncaged/workflow-template-develop` | `src/moderator.ts` | `developTable` — conditional multi-phase table |
-
-## See Also
-
- [Bundle](./bundle.md) — the build artifact produced from a template + adapter
- [Role](./role.md) — the `RoleDefinition` type each template role implements
- [Engine](./engine.md) — the execution loop that drives the template's `WorkflowFn`
@@ -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
@@ -9,3 +9,5 @@ bunfig.toml
 xiaoju/
 solve-issue-entry.ts
 packages/workflow-template-develop/develop.esm.js
+.DS_Store
+*.py
@@ -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 | ...
 ```
@@ -1,71 +1,93 @@
 # @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
+## Package Map

-| 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. |
+| Package | npm | Role |
+|---------|-----|------|
+| `cli-workflow` | `@uncaged/cli-workflow` | `uwf` CLI binary — thread lifecycle, workflow registry, CAS inspection, setup |
+| `workflow-protocol` | `@uncaged/workflow-protocol` | Shared TypeScript types (`WorkflowPayload`, `StepNodePayload`, `WorkflowConfig`, etc.) |
+| `workflow-moderator` | `@uncaged/workflow-moderator` | JSONata graph evaluator — determines next role or `$END` |
+| `workflow-agent-kit` | `@uncaged/workflow-agent-kit` | `createAgent` factory, context builder, two-layer extract pipeline |
+| `workflow-agent-hermes` | `@uncaged/workflow-agent-hermes` | `uwf-hermes` agent — spawns Hermes chat, captures session |
+| `workflow-util` | `@uncaged/workflow-util` | Crockford Base32, ULID, logger, frontmatter parsing |

-## Monorepo Packages
-
-```
-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)
-```
-
-Managed with **bun workspace** using the `workspace:*` protocol.
+External: [`@uncaged/json-cas`](https://www.npmjs.com/package/@uncaged/json-cas) (CAS store + JSON Schema validation) + `@uncaged/json-cas-fs` (filesystem backend).

 ## Quick Start

 ```bash
-# Install dependencies
-bun install
+# 1. Configure provider and model
+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
+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
+## CLI Commands

-```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
-```
+### Thread

-Run `uncaged-workflow` with no arguments to print usage, or `uncaged-workflow skill cli` for the full CLI skill reference.
+| Command | Description |
+|---------|-------------|
+| `uwf thread start <workflow> -p <prompt>` | Create a thread (no execution) |
+| `uwf thread step <thread-id> [--agent <cmd>]` | Execute one moderator→agent→extract cycle |
+| `uwf thread show <thread-id>` | Show head pointer and done status |
+| `uwf thread list [--all]` | List threads (`--all` includes archived) |
+| `uwf thread steps <thread-id>` | List all steps chronologically |
+| `uwf thread read <thread-id> [--quota N]` | 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 |
+| `uwf thread kill <thread-id>` | Terminate and archive |
+
+### 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>` | Read a CAS node |
+| `uwf cas put <type-hash> <data>` | Store a node |
+| `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 schemas |
+| `uwf cas schema get <hash>` | Show a schema |
+
+### Setup
+
+| Command | Description |
+|---------|-------------|
+| `uwf setup` | Interactive provider/model/agent configuration |
+| `uwf setup --provider ... --base-url ... --api-key ... --model ...` | Non-interactive setup |
+
+Config stored in `~/.uncaged/workflow/config.yaml`. API keys in `~/.uncaged/workflow/.env`.

 ## 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 check              # tsc + biome + lint-log-tags
+bun run format             # Auto-format with Biome
+bun test                   # Run all tests
 ```

+Managed with **bun workspace**. See [CLAUDE.md](CLAUDE.md) for coding conventions.
+
 ## Architecture

-See [docs/architecture.md](docs/architecture.md) for the full design — three-phase engine loop, bundle contract, storage layout, and design decisions.
+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.
@@ -1,16 +0,0 @@
-import { createCursorAgent } from "./packages/workflow-agent-cursor/src/index.js";
-import { createWorkflow } from "./packages/workflow-runtime/src/create-workflow.js";
-import {
-  buildDevelopDescriptor,
-  developWorkflowDefinition,
-} from "./packages/workflow-template-develop/src/index.js";
-
-const agent = createCursorAgent({
-  command: "/home/azureuser/.local/bin/cursor-agent",
-  model: "auto",
-  timeout: 300_000,
-  workspace: null,
-});
-
-export const descriptor = buildDevelopDescriptor();
-export const run = createWorkflow(developWorkflowDefinition, { adapter: agent, overrides: null });
@@ -1,268 +1,495 @@
-# Uncaged workflow — Architecture
+# Workflow Engine — Architecture

-**Last updated:** 2026-05-09
+**Last updated:** 2026-05-19

 ---

 ## Overview

-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). No daemon — processes start on demand and exit when done.
+A stateless workflow engine driven by a single-step CLI. 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.

-The implementation lives in **15** Bun workspace packages under `packages/`, using the `workspace:*` protocol.
+The implementation lives in **6** active packages under `packages/`, plus two external CAS packages (`@uncaged/json-cas`, `@uncaged/json-cas-fs`). Legacy packages reside in `legacy-packages/` and are not part of the active stack.

 ## Package map

-Grouped by responsibility (npm name → folder).
-
 | Layer | Package | One-line role |
-|-------|---------|----------------|
-| Contract | `@uncaged/workflow-protocol` → `workflow-protocol` | Shared TypeScript types and `Result` helpers; peer `zod` only — no other workspace deps. |
-| Author API | `@uncaged/workflow-runtime` → `workflow-runtime` | `createWorkflow` and re-exports of protocol workflow types for bundle authors. |
-| Shared infra | `@uncaged/workflow-util` → `workflow-util` | Base32/ULID, logger, storage root paths, global CAS dir, ref-field helpers. |
-| LLM plumbing | `@uncaged/workflow-reactor` → `workflow-reactor` | `createLlmFn`, `createThreadReactor`, and related tool-call types for threaded LLM invocation. |
-| CAS | `@uncaged/workflow-cas` → `workflow-cas` | `CasStore` implementation, XXH64 hashing, Merkle helpers over CAS payloads. |
-| Registry / bundles | `@uncaged/workflow-register` → `workflow-register` | Bundle validation & dynamic export extraction, `workflow.yaml` registry I/O, provider/model resolution. |
-| Engine | `@uncaged/workflow-execute` → `workflow-execute` | Thread execution, worker entry path, fork/GC, extract pipeline, `workflowAsAgent`. |
-| CLI | `@uncaged/cli-workflow` → `cli-workflow` | `uncaged-workflow` binary (depends on engine, registry, CAS, protocol, util, runtime). |
-| Agent adapters | `@uncaged/workflow-agent-cursor` → `workflow-agent-cursor` | `AgentFn` via `cursor-agent` CLI + workspace extraction. |
-| | `@uncaged/workflow-agent-hermes` → `workflow-agent-hermes` | `AgentFn` via `hermes chat` CLI. |
-| | `@uncaged/workflow-agent-llm` → `workflow-agent-llm` | `AgentFn` via OpenAI-compatible HTTP (`LlmProvider` from runtime). |
-| Agent shared | `@uncaged/workflow-util-agent` → `workflow-util-agent` | `buildAgentPrompt`, `spawnCli` for CLI-backed agents. |
-| Templates | `@uncaged/workflow-template-develop` → `workflow-template-develop` | Develop workflow definition, roles, descriptor builder. |
-| | `@uncaged/workflow-template-solve-issue` → `workflow-template-solve-issue` | Solve-issue workflow definition, roles, descriptor builder. |
-| Dashboard | `@uncaged/workflow-dashboard` → `workflow-dashboard` | Private Vite + React app (`src/main.tsx`); only `react` / `react-dom` dependencies — no workspace packages. |
+|-------|---------|---------------|
+| Contract | `@uncaged/workflow-protocol` → `workflow-protocol` | Shared TypeScript types (`WorkflowPayload`, `StepNodePayload`, `ModeratorContext`, `WorkflowConfig`, etc.). No runtime deps beyond `@uncaged/json-cas-fs`. |
+| Shared infra | `@uncaged/workflow-util` → `workflow-util` | Crockford Base32, ULID generation, `createLogger`, frontmatter parsing/validation. |
+| Moderator | `@uncaged/workflow-moderator` → `workflow-moderator` | JSONata-based graph evaluator: given a `WorkflowPayload` and `ModeratorContext`, returns the next role or `$END`. |
+| Agent framework | `@uncaged/workflow-agent-kit` → `workflow-agent-kit` | `createAgent` entrypoint factory, context builder, frontmatter fast-path extractor, LLM extract fallback, output format instruction builder. |
+| Agent: Hermes | `@uncaged/workflow-agent-hermes` → `workflow-agent-hermes` | `uwf-hermes` CLI binary — spawns `hermes chat`, pipes prompt, captures session detail. |
+| CLI | `@uncaged/cli-workflow` → `cli-workflow` | `uwf` binary — thread lifecycle, workflow registry, CAS inspection, setup. |

-## Dependency graph (workspace packages)
+### External dependencies

-Bottom-up layering for the execution stack:
+| Package | Role |
+|---------|------|
+| `@uncaged/json-cas` | Content-addressed store API, XXH64 hashing, JSON Schema registration and validation. |
+| `@uncaged/json-cas-fs` | Filesystem backend for `json-cas`. |
+| `jsonata` | JSONata expression evaluator (used by `workflow-moderator`). |
+| `commander` | CLI argument parsing (used by `cli-workflow`). |
+| `dotenv` | Loads `.env` files for API keys. |
+| `yaml` | YAML parse/stringify. |
+
+## Dependency graph

 ```mermaid
 flowchart BT
+  subgraph External
+    jcas["@uncaged/json-cas"]
+    jcasfs["@uncaged/json-cas-fs"]
+  end
  subgraph L0["Layer 0 — contract"]
    protocol["@uncaged/workflow-protocol"]
  end
-  subgraph L1["Layer 1 — on protocol"]
-    runtime["@uncaged/workflow-runtime"]
+  subgraph L1["Layer 1 — shared"]
    util["@uncaged/workflow-util"]
-    reactor["@uncaged/workflow-reactor"]
+    moderator["@uncaged/workflow-moderator"]
  end
-  subgraph L2["Layer 2 — protocol + util"]
-    cas["@uncaged/workflow-cas"]
-    register["@uncaged/workflow-register"]
+  subgraph L2["Layer 2 — agent framework"]
+    kit["@uncaged/workflow-agent-kit"]
  end
-  subgraph L3["Layer 3 — engine"]
-    execute["@uncaged/workflow-execute"]
+  subgraph L3["Layer 3 — agent implementations"]
+    hermes["@uncaged/workflow-agent-hermes"]
  end
  subgraph L4["Layer 4 — CLI"]
    cli["@uncaged/cli-workflow"]
  end
-  runtime --> protocol
+  protocol --> jcasfs
  util --> protocol
-  reactor --> protocol
-  cas --> protocol
-  cas --> util
-  register --> protocol
-  register --> util
-  execute --> protocol
-  execute --> runtime
-  execute --> util
-  execute --> cas
-  execute --> reactor
-  execute --> register
+  moderator --> protocol
+  kit --> protocol
+  kit --> util
+  kit --> jcas
+  kit --> jcasfs
+  hermes --> kit
+  hermes --> jcas
  cli --> protocol
  cli --> util
-  cli --> cas
-  cli --> execute
-  cli --> register
-  cli --> runtime
+  cli --> kit
+  cli --> moderator
+  cli --> jcas
+  cli --> jcasfs
 ```

-**Adjacent consumers** (not in the main CLI stack):
+## Workflow definition

- `@uncaged/workflow-util-agent` → `@uncaged/workflow-runtime`
- `@uncaged/workflow-agent-llm` → `@uncaged/workflow-runtime`
- `@uncaged/workflow-agent-cursor` → `@uncaged/workflow-runtime`, `@uncaged/workflow-util-agent`, `zod`
- `@uncaged/workflow-agent-hermes` → `@uncaged/workflow-runtime`, `@uncaged/workflow-util-agent`
- `@uncaged/workflow-template-develop` → `@uncaged/workflow-register`, `@uncaged/workflow-runtime`, `zod`
- `@uncaged/workflow-template-solve-issue` → `@uncaged/workflow-register`, `@uncaged/workflow-runtime`, `zod` (dev-only workspace deps: `@uncaged/workflow-cas`, `@uncaged/workflow-execute` for tests/tooling per `package.json`)
+Workflows are **YAML files** (not ESM bundles). `uwf workflow put <file.yaml>` parses the YAML, registers output schemas as JSON Schema CAS nodes, and stores the `WorkflowPayload` as a CAS node.

-## Package roles (detail)
+Example (`examples/solve-issue.yaml`):

- **`workflow-protocol`** — Pure types (`WorkflowFn`, contexts, `CasStore` interface, descriptor shapes), `START` / `END`, `ok` / `err`. Depends only on peer `zod` for schema-related types in signatures.
- **`workflow-runtime`** — Workflow author surface: `createWorkflow` from `src/create-workflow.js`, re-exports protocol types/constants used when authoring bundles.
- **`workflow-util`** — Cross-cutting utilities: Crockford Base32, ULID, `createLogger`, `getDefaultWorkflowStorageRoot`, `getGlobalCasDir`, ref normalization; re-exports `ok`/`err` from protocol.
- **`workflow-cas`** — Filesystem CAS (`createCasStore`), `hashString` / `hashWorkflowBundleBytes`, Merkle node serialization and helpers (`merkle.js`).
- **`workflow-register`** — Bundle pipeline (`validateWorkflowBundle`, `extractBundleExports`, descriptor builders), registry YAML read/write, `resolveModel` / `splitProviderModelRef`.
- **`workflow-execute`** — `executeThread`, supervisor/worker wiring (`engine/`), fork/GC/pause gate, `createExtract` + LLM extract helpers (`extract/`), `workflowAsAgent`. Imports `@uncaged/workflow-reactor` for LLM-backed extract/supervisor paths (`extract-fn.ts`, `supervisor.ts`).
- **`workflow-reactor`** — `createLlmFn`, `createThreadReactor`, and thread tool-invocation types — consumed by `workflow-execute`.
- **`cli-workflow`** — CLI commands and HTTP/dashboard-related wiring (`hono`, `yaml`); composes register + execute + CAS + util.
- **`workflow-agent-*`** — Replaceable `AgentFn` implementations (Cursor / Hermes CLIs, or HTTP LLM).
- **`workflow-util-agent`** — Shared prompt assembly and subprocess spawning for CLI agents.
- **`workflow-template-*`** — Concrete `WorkflowDefinition` graphs + Zod role schemas + descriptor builders for publishing bundles.
- **`workflow-dashboard`** — Standalone React UI; no published library entry matching `src/index.ts`.
+```yaml
+name: "solve-issue"
+description: "End-to-end issue resolution"
+roles:
+  planner:
+    description: "Creates implementation plan"
+    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 }
+        steps: { type: array, items: { type: string } }
+      required: [plan, steps]
+  developer:
+    description: "Implements code changes"
+    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 } }
+        summary: { type: string }
+      required: [filesChanged, summary]
+  reviewer:
+    description: "Reviews code changes"
+    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 }
+        comments: { type: string }
+      required: [approved, comments]
+conditions:
+  notApproved:
+    description: "Reviewer rejected the implementation"
+    expression: "steps[-1].output.approved = false"
+graph:
+  $START:
+    - role: "planner"
+      condition: null
+  planner:
+    - role: "developer"
+      condition: null
+  developer:
+    - role: "reviewer"
+      condition: null
+  reviewer:
+    - role: "developer"
+      condition: "notApproved"
+    - role: "$END"
+      condition: null
+```
+
+Key properties:
+
+- **`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`
+- **No Zod** — all schemas are JSON Schema, validated through `@uncaged/json-cas`

 ## Three-phase engine loop

-Each role round is implemented in `packages/workflow-runtime/src/create-workflow.ts` (`advanceOneRound`): moderator → agent → extractor, with progressive context types from `@uncaged/workflow-protocol`.
+Each `uwf thread step` runs exactly one cycle: moderator → agent → extract. The CLI orchestrates this in `packages/cli-workflow/src/commands/thread.ts` (`cmdThreadStep`).

 ```
 ┌─→ Phase 1: MODERATOR
-│   Context: ModeratorContext { threadId, depth, start, steps }
-│   Action:  moderator(ctx) → role name | END
+│   Input:  WorkflowPayload + ModeratorContext { start, steps[] }
+│   Engine: JSONata conditions evaluated against the graph
+│   Output: next role name | $END
 │
 │   Phase 2: AGENT
-│   Context: AgentContext = ModeratorCtx + { currentRole: { name, systemPrompt } }
-│   Action:  agent(ctx) → raw string
+│   Input:  thread-id + role (via argv)
+│   Engine: agent-kit builds context from CAS chain, prepends
+│           output format instruction to system prompt, spawns agent
+│   Output: raw string (frontmatter markdown)
 │
-│   Phase 3: EXTRACTOR
-│   Context: ExtractContext = AgentCtx + { agentContent }
-│   Action:  runtime.extract(schema, extractPrompt, ctx) → typed meta
+│   Phase 3: EXTRACT
+│   Input:  raw agent output + role's meta schema
+│   Engine: two-layer extract (frontmatter fast path → LLM fallback)
+│   Output: CasRef to structured output node
 │
-│   Merge: RoleStep { role, contentHash, meta, refs, timestamp }
-│   Append to steps
-└─────────────────────────────────────────────────────┘
+│   Persist: StepNode { start, prev, role, output, detail, agent }
+│   Update:  threads.yaml head pointer
+└─────────────────────────────────────────────────────────────────┘
 ```

-### Context types (progressive)
+### Context types

 Defined in `packages/workflow-protocol/src/types.ts`:

 ```typescript
-type ModeratorContext<M> = ThreadContext<M>;
-type AgentContext<M> = ModeratorContext<M> & {
-  currentRole: { name: string; systemPrompt: string };
+type StepContext = {
+  role: string;
+  output: unknown;    // CAS node payload, expanded (not hash)
+  detail: CasRef;
+  agent: string;
+};
+
+type ModeratorContext = {
+  start: StartNodePayload;  // { workflow: CasRef, prompt: string }
+  steps: StepContext[];     // chronological, oldest first
+};
+
+type AgentContext = ModeratorContext & {
+  threadId: ThreadId;
+  role: string;
+  store: Store;
+  workflow: WorkflowPayload;
+  outputFormatInstruction: string;
 };
-type ExtractContext<M> = AgentContext<M> & { agentContent: string };
 ```

 ### Key properties

- **Moderator is synchronous and pure** — no I/O, no state mutation inside `createWorkflow`’s moderator call path.
- **Agent receives `AgentContext`** — reads `ctx.currentRole.systemPrompt`; raw output becomes `agentContent` for extract.
- **Extractor is `WorkflowRuntime.extract`** — supplied by the engine from registry-resolved LLM config (`workflow-execute`); stores agent body in CAS and yields `contentHash` + `refs` on each step (`create-workflow.ts`).
- **`extractPrompt` is a call parameter** on `RoleDefinition`, not implicit context state.
+- **Moderator** — pure JSONata evaluation; no LLM call, no I/O beyond CAS reads. Evaluates `workflow.graph[currentRole]` transitions in order, returns first match.
+- **Agent** — receives `AgentContext` with thread history + role system prompt + output format instruction. Raw output is frontmatter markdown.
+- **Extractor** — two-layer: tries frontmatter fast-path first (zero LLM cost), falls back to LLM extract if frontmatter is absent or invalid.
+- **Stateless** — each `uwf thread step` is an atomic, self-contained operation. No in-memory state between steps.

-## Agent information sources
+## Agent CLI protocol

-An agent has exactly three information sources:
+Each agent is an external command invoked by `uwf thread step`:

-1. **Prior knowledge** — LLM training, agent memory, agent skills
-2. **Thread context** — `AgentContext` (`start`, `steps`, `currentRole`)
-3. **Derived information** — from 1 & 2 (e.g. tool calls, shell commands)
-
-No hidden environment parameters. If an agent needs something (like a workspace path), it obtains it via `ExtractFn` (e.g. Cursor agent).
-
-## Bundle contract
-
-A workflow bundle is a single `.esm.js` file with two named exports (see `WorkflowFn` / `WorkflowDescriptor` in `packages/workflow-protocol/src/types.ts`):
-
-```typescript
-export const descriptor: WorkflowDescriptor;
-export const run: WorkflowFn;
-
-type WorkflowFn = (
-  thread: ThreadContext,
-  runtime: WorkflowRuntime,
-) => AsyncGenerator<RoleOutput, WorkflowCompletion>;
+```bash
+<agent-cmd> <thread-id> <role>
 ```

-`RoleOutput` carries `contentHash`, `meta`, and `refs` (agent text lives in CAS, addressed by hash).
+Contract:
+1. `uwf thread step` determines the next role via the moderator
+2. Agent CLI is spawned with `(thread-id, role)` as positional args
+3. `workflow-agent-kit` (`createAgent`) handles the boilerplate:
+   - Parses argv
+   - Loads `.env` from storage root
+   - Builds `AgentContext` by walking the CAS chain from `threads.yaml` head
+   - 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)
+   - Prints the new `StepNode` CAS hash to stdout
+4. `uwf thread step` reads stdout, updates `threads.yaml` head pointer, re-evaluates moderator for `done`
+5. Exit 0 = success, non-zero = failure

-### Constraints
+Agent resolution priority: `--agent` CLI override → `config.yaml` per-workflow/role override → `config.yaml` `defaultAgent`.

- Single `.esm.js` file
- No dynamic `import()` in bundles (loader exempt in engine)
- Portable bundle static imports are constrained by validation in `@uncaged/workflow-register` (`validateWorkflowBundle`)
- XXH64 hash (Crockford Base32) = version ID
+## Agent output format: frontmatter markdown (RFC #351)

-### Why AsyncGenerator?
+Agents produce **frontmatter markdown** — YAML frontmatter for structured meta, followed by a markdown body for content:

- Each `yield` lets `workflow-execute` persist state, CAS rows, and enforce pause/abort
- `return` supplies `WorkflowCompletion`
- Fork replays historical steps into a new thread context
- Bundle does not import the engine — only protocol/runtime types at build time
+```markdown
+---
+status: done
+next: reviewer
+confidence: 0.9
+artifacts:
+  - src/auth.ts
+scope: role
+---
+
+## Implementation
+
+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 `meta` JSON Schema.
+
+## Two-layer extract
+
+Structured output extraction uses a two-layer strategy (`workflow-agent-kit`):
+
+### Layer 1: frontmatter fast path (`frontmatter.ts`)
+
+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 `meta` schema
+5. Validate with `json-cas` schema validation
+6. If valid → return `outputHash` (zero LLM cost)
+
+### Layer 2: LLM extract fallback (`extract.ts`)
+
+If the fast path returns `null` (no frontmatter, invalid, or doesn't satisfy schema):
+
+1. Resolve extract model alias from config (`modelOverrides.extract` → `models.extract` → `defaultModel`)
+2. Call OpenAI-compatible chat completion with JSON mode
+3. System prompt: "Extract structured data matching this JSON Schema: ..."
+4. User message: the raw agent output
+5. Parse response, `store.put()`, validate
+6. Return `outputHash`
+
+## Prompt injection
+
+`workflow-agent-kit` prepends two pieces of context to the agent's system prompt:
+
+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.
+
+## CAS node types
+
+### Workflow
+
+```yaml
+type: <workflow-schema-hash>
+payload:
+  name: "solve-issue"
+  description: "End-to-end issue resolution"
+  roles:
+    planner:
+      description: "Creates implementation plan"
+      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"
+      expression: "steps[-1].output.approved = false"
+  graph:
+    $START:
+      - role: "planner"
+        condition: null
+```
+
+### StartNode
+
+```yaml
+type: <start-node-schema-hash>
+payload:
+  workflow: "4KNM2PXR3B1QW"    # cas_ref → Workflow
+  prompt: "Fix the login bug..."
+```
+
+### StepNode
+
+```yaml
+type: <step-node-schema-hash>
+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 meta schema)
+  detail: "7BQST3VW9F2MA"     # cas_ref → execution detail (raw turns, session data)
+  agent: "uwf-hermes"         # agent command used (plain string)
+```
+
+### Chain structure
+
+```
+threads.yaml: { "01J7K9...4T": "8FWKR3TN5V1QA" }
+                                    │
+                                    ▼
+                            StepNode (step 3)
+                            ├── start ──→ StartNode
+                            │              ├── workflow → Workflow (CAS)
+                            │              └── prompt: "Fix..."
+                            ├── prev ──→ StepNode (step 2)
+                            │             ├── prev ──→ StepNode (step 1)
+                            │             │             └── prev: null
+                            │             └── ...
+                            ├── role: "reviewer"
+                            ├── output → CAS({ approved: true })
+                            ├── detail → CAS(session turns)
+                            └── agent: "uwf-hermes"
+```

 ## Storage layout

 ```
 ~/.uncaged/workflow/
-├── cas/                           # Global content-addressed blobs (see getGlobalCasDir)
-├── bundles/
-│   ├── C9NMV6V2TQT81.esm.js       # Crockford Base32 of XXH64
-│   ├── C9NMV6V2TQT81.yaml         # Role descriptor sidecar (when present)
-│   └── C9NMV6V2TQT81/             # Per-hash bundle dir (alongside or instead of loose files)
-│       ├── threads.json           # Active threads: threadId → { head, start, updatedAt }
-│       └── history/
-│           └── 2026-05-09.jsonl   # Completed threads (one JSON object per line)
-├── logs/                          # One folder per bundle hash
-│   └── C9NMV6V2TQT81/
-│       ├── 01KQXKW…YG.running     # Present while worker executes this thread (optional)
-│       └── 01KQXKW…YG.info.jsonl   # Debug log
-└── workflow.yaml                  # Registry
+├── cas/                          # json-cas filesystem store (all CAS nodes)
+├── config.yaml                   # Provider, model, agent configuration
+├── threads.yaml                  # Active thread head pointers: threadId → CasRef
+├── history.jsonl                 # Archived thread records
+├── registry.yaml                 # Workflow name → CAS hash mapping
+└── .env                          # API keys (loaded by dotenv)
 ```

+### Mutable state
+
+Only three files carry mutable state:
+
+| File | Contents |
+|------|----------|
+| `threads.yaml` | `Record<ThreadId, CasRef>` — maps active thread IDs to head node hash |
+| `history.jsonl` | Append-only log of completed threads (`thread`, `workflow`, `head`, `completedAt`) |
+| `registry.yaml` | Workflow name → current CAS hash |
+
+Everything else is immutable CAS content.
+
 ### ID encoding: Crockford Base32

 - Case-insensitive, filesystem-safe, no ambiguous chars (0/O, 1/I/L)
- Bundle hash: XXH64 → 13-char
- Thread ID: ULID → 26-char (10 timestamp + 16 random)
+- CAS hash: XXH64 → 13-char Crockford Base32
+- Thread ID: ULID → 26-char Crockford Base32 (10 timestamp + 16 random)

-### Registry (`workflow.yaml`)
+### Config (`config.yaml`)

-Managed by `@uncaged/workflow-register` (`readWorkflowRegistry`, `writeWorkflowRegistry`, …). Shape includes workflow entries and a top-level `config` section used for extract/supervisor model resolution.
+```yaml
+providers:
+  openrouter:
+    baseUrl: "https://openrouter.ai/api/v1"
+    apiKeyEnv: "OPENROUTER_API_KEY"

-### Thread storage (CAS + index)
+models:
+  sonnet:
+    provider: "openrouter"
+    name: "anthropic/claude-sonnet-4"
+  gpt4o-mini:
+    provider: "openai"
+    name: "gpt-4o-mini"

-Thread execution state is a chain of immutable CAS nodes (`StartNode`, `StateNode`, content Merkle blobs). Per bundle:
+agents:
+  hermes:
+    command: "uwf-hermes"
+    args: []
+  cursor:
+    command: "uwf-cursor"
+    args: []

- **`threads.json`** — only in-flight threads (`head`, `start`, `updatedAt`).
- **`history/{YYYY-MM-DD}.jsonl`** — completed threads (`threadId`, `head`, `start`, `completedAt`).
- **CAS (`cas/`)** — payloads and refs for replay, GC, and fork sharing.
+defaultAgent: "hermes"
+agentOverrides:
+  solve-issue:
+    developer: "cursor"

-**`.info.jsonl`** — Structured debug log via `@uncaged/workflow-util` `createLogger`:
-
-```jsonc
-{ "tag": "4KNMR2PX", "content": "Loading bundle...", "timestamp": ... }
+defaultModel: "sonnet"
+modelOverrides:
+  extract: "gpt4o-mini"
 ```

-Tags are 8-char Crockford Base32 (40-bit random), one per call site. `grep "4KNMR2PX"` → code location.
-
-## Execution model
-
- **No daemon.** `uncaged-workflow run <name>` starts a worker process (`workflow-execute` worker entry via `getWorkerHostScriptPath`)
- Threads share bundle-scoped workers as implemented in CLI/engine
- Pause/resume/abort via engine IPC and pause gate (`createThreadPauseGate`)
-
 ## CLI commands

-| Priority | Command | Description |
-|----------|---------|-------------|
-| P1 | `add <name> <file.esm.js>` | Register a bundle |
-| P1 | `list` | List registered workflows |
-| P1 | `show <name>` | Show workflow details |
-| P1 | `remove <name>` | Remove a workflow |
-| P1 | `run <name> [--prompt] [--max-rounds]` | Start a thread |
-| P1 | `threads [name]` | List threads |
-| P1 | `thread <id>` | Show thread state |
-| P1 | `thread rm <id>` | Delete a thread |
-| P1 | `ps` | List running threads |
-| P1 | `kill <thread-id>` | Terminate a running thread |
-| P2 | `history <name>` | Show version history |
-| P2 | `rollback <name> [hash]` | Switch to a previous version |
-| P2 | `pause <thread-id>` | Pause a running thread |
-| P2 | `resume <thread-id>` | Resume a paused thread |
-| P3 | `fork <thread-id> [--from-role <role>]` | Fork from historical state |
+Binary: `uwf`
+
+### Thread commands
+
+| Command | Description |
+|---------|-------------|
+| `uwf thread start <workflow> -p <prompt>` | Create a thread (StartNode → CAS, head → threads.yaml). No execution. |
+| `uwf thread step <thread-id> [--agent <cmd>]` | Execute one moderator→agent→extract cycle. |
+| `uwf thread show <thread-id>` | Show thread head pointer and done status. |
+| `uwf thread list [--all]` | List active threads (`--all` includes archived). |
+| `uwf thread steps <thread-id>` | List all steps in chronological order. |
+| `uwf thread read <thread-id> [--quota <chars>] [--before <hash>]` | Render thread as human-readable markdown. |
+| `uwf thread fork <step-hash>` | Fork a thread from a specific CAS node. |
+| `uwf thread step-details <step-hash>` | Dump full detail node as YAML. |
+| `uwf thread kill <thread-id>` | Terminate and archive a thread. |
+
+### Workflow commands
+
+| Command | Description |
+|---------|-------------|
+| `uwf workflow put <file.yaml>` | Register a workflow from YAML definition. |
+| `uwf workflow show <id>` | Show workflow by name or CAS hash. |
+| `uwf workflow list` | List registered workflows. |
+
+### CAS commands
+
+| Command | Description |
+|---------|-------------|
+| `uwf cas get <hash>` | Read a CAS node. |
+| `uwf cas put <type-hash> <data>` | Store a node, print its hash. |
+| `uwf cas has <hash>` | Check if a hash exists. |
+| `uwf cas refs <hash>` | List direct CAS references. |
+| `uwf cas walk <hash>` | Recursive traversal from a node. |
+| `uwf cas reindex` | Rebuild type index from all nodes. |
+| `uwf cas schema list` | List registered schemas. |
+| `uwf cas schema get <hash>` | Show a schema by type hash. |
+
+### Setup
+
+| Command | Description |
+|---------|-------------|
+| `uwf setup [--provider --base-url --api-key --model --agent]` | Configure provider/model/agent (interactive if no flags). |
+
+## Toolchain
+
+| Tool | Purpose |
+|------|---------|
+| **bun** | Package manager + runtime |
+| **TypeScript** | Type checking (strict mode) |
+| **Biome** | Lint + format |
+| **vitest** | Test runner |

 ## Design decisions

 | Decision | Rationale |
 |----------|-----------|
-| **Role = pure data** | Decouples definition from execution; same role with different agents |
-| **Agent bound at runtime** | `WorkflowDefinition` is reusable; agent choice is deployment concern |
-| **Three-phase context** | Each phase sees only what it needs; types live in `workflow-protocol` |
-| **`WorkflowRuntime.extract` + CAS `contentHash`** | Large agent bodies deduplicated globally; Merkle roots summarize threads |
-| **`workflow-reactor` split** | LLM tool-calling loop isolated from filesystem/registry concerns |
-| **Single-file ESM** | Hash = version, self-contained bundle |
-| **No daemon** | OS handles process lifecycle |
-| **Crockford Base32** | Filesystem-safe, readable, compact |
-| **15-package split** | Clear boundaries: protocol ↔ runtime author API ↔ util/CAS/register ↔ execute ↔ CLI ↔ agents/templates/UI |
+| **YAML workflow definitions** | Human-readable, versionable, no build step required. JSON Schema inline in YAML, registered as CAS nodes on `workflow put`. |
+| **Stateless single-step CLI** | Each `uwf thread step` is atomic — no in-memory state, no daemon, no long-running process. OS handles lifecycle. |
+| **CAS-backed thread state** | Immutable linked nodes enable fork, replay, and GC without copying data. Content-addressed deduplication across threads. |
+| **JSONata moderator** | Declarative condition expressions evaluated against thread history. No LLM cost for routing decisions. |
+| **Frontmatter markdown output** | Agents produce structured meta (YAML frontmatter) alongside free-form content (markdown body). Enables zero-cost extraction when frontmatter is well-formed. |
+| **Two-layer extract** | Fast path avoids LLM calls when agents follow the format; LLM fallback handles messy output gracefully. |
+| **Prompt injection for format** | Output format instruction prepended to system prompt ensures agents produce parseable output without per-agent configuration. |
+| **JSON Schema (not Zod)** | Schemas are CAS-native data — storable, hashable, validatable through `json-cas`. No code generation, no runtime library dependency. |
+| **Agent as external command** | Agents are independent CLI binaries (`uwf-hermes`, `uwf-cursor`). Swappable per workflow/role via config. No tight coupling to the engine. |
+| **No daemon** | Process starts, does one step, exits. Simpler failure model, no connection management. |
+| **Crockford Base32** | Filesystem-safe, case-insensitive, readable, compact. |
@@ -0,0 +1,387 @@
+# 设计文档：office-agent 文档生成/编辑 Workflow 体系
+
+**日期：** 2026-05-18
+
+---
+
+## 概述
+
+在 monorepo 中新增三个包，实现通过 `office-agent` CLI 生成或编辑 Word 文档的完整 workflow 体系。
+
+| 包 | npm name | 职责 |
+|---|---|---|
+| `workflow-template-document` | `@uncaged/workflow-template-document` | 纯结构：角色定义、meta schema、调度表、descriptor |
+| `workflow-agent-office` | `@uncaged/workflow-agent-office` | writer 角色执行器：调用 `office-agent` CLI |
+| `workflow-agent-docx-diff` | `@uncaged/workflow-agent-docx-diff` | differ 角色执行器：调用 `docx-diff` CLI |
+
+Template 只定义结构，不含执行逻辑。执行器与 template 解耦。
+
+---
+
+## 一、`workflow-template-document`
+
+### Thread 启动输入
+
+```typescript
+// src/types.ts
+type DocumentStartInput = {
+  prompt: string;           // 用户指令
+  inputDocx: string | null; // null = 生成模式；本机绝对路径 = 编辑模式
+};
+```
+
+start.content 为 JSON `{ prompt, inputDocx }` 或纯文本（fallback：generate 模式，整段作为 prompt）。
+
+### 角色与 Meta
+
+`WriterMeta` 使用 discriminated union，在 schema 层区分两种模式：
+
+```typescript
+const writerMetaSchema = z.discriminatedUnion("mode", [
+  z.object({
+    mode: z.literal("generate"),
+    outputDocx: z.string(),   // 生成产物绝对路径
+    sourceDocx: z.null(),
+  }),
+  z.object({
+    mode: z.literal("edit"),
+    outputDocx: z.string(),   // 修改后产物：<outputDir>/modified.docx
+    sourceDocx: z.string(),   // 原始副本：<outputDir>/original.docx
+  }),
+]);
+type WriterMeta = z.infer<typeof writerMetaSchema>;
+
+// differ：仅编辑模式执行
+const differMetaSchema = z.object({
+  sourceDocx: z.string(),
+  modifiedDocx: z.string(),
+  diffDocx: z.string(),
+});
+type DifferMeta = z.infer<typeof differMetaSchema>;
+```
+
+两个角色的 `systemPrompt` 均为 `""`。
+
+### 调度表
+
+```
+START → writer ──(mode = "edit")──→ differ → END
+               ↘(mode = "generate")→ END
+```
+
+### 公开导出
+
+template 导出两个对象供消费方使用：
+
+- `documentWorkflowDefinition: WorkflowDefinition<DocumentMeta>` — 传入 `createWorkflow` 的 `def` 参数
+- `buildDocumentDescriptor(): WorkflowDescriptor` — bundle 导出用
+
+```typescript
+// bundle 侧用法
+export const descriptor = buildDocumentDescriptor();
+export const run = createWorkflow(documentWorkflowDefinition, { adapter, overrides });
+```
+
+### 包文件结构
+
+```
+packages/workflow-template-document/
+  src/
+    types.ts           # DocumentStartInput
+    roles/
+      writer.ts        # writerMetaSchema, WriterMeta, writerRole
+      differ.ts        # differMetaSchema, DifferMeta, differRole
+      index.ts
+    roles.ts           # DocumentMeta, documentRoles
+    moderator.ts       # writerIsEditMode condition + documentTable
+    definition.ts      # documentWorkflowDefinition
+    descriptor.ts      # buildDocumentDescriptor()
+    index.ts
+  __tests__/
+    moderator.test.ts
+  package.json
+  tsconfig.json
+```
+
+### 依赖
+
+```json
+{
+  "@uncaged/workflow-protocol": "workspace:^",
+  "@uncaged/workflow-runtime": "workspace:^",
+  "@uncaged/workflow-register": "workspace:^",
+  "zod": "^4.0.0"
+}
+```
+
+---
+
+## 二、`workflow-agent-office`
+
+### office-agent CLI 接口
+
+```bash
+# 生成模式：在 CWD 生成 output.docx
+office-agent create "<prompt>" -o output.docx
+
+# 编辑模式：在 CWD 对 modified.docx 进行修改（覆写）
+office-agent edit modified.docx "<instruction>"
+```
+
+- 两个命令均为阻塞调用（CLI 内部消费 SSE，退出即完成）
+- 输出文件落到调用方设定的 CWD
+- 退出码 0 = 成功，非零 = 失败
+
+### 文件命名约定
+
+| 模式 | 文件 | 路径 |
+|---|---|---|
+| generate | 输出 | `<outputDir>/output.docx` |
+| edit | 原始副本（workflow-owned 快照） | `<outputDir>/original.docx` |
+| edit | 修改后产物 | `<outputDir>/modified.docx` |
+
+edit 模式先将 `inputDocx` 复制为 `original.docx`（不可变快照），再复制为 `modified.docx`，对 `modified.docx` 调用 CLI。agent 覆写 `modified.docx`，`original.docx` 保持不变。differ 对比这两个 workflow-owned 文件，不依赖用户原始路径。
+
+### 执行流程
+
+**生成模式（`inputDocx = null`）：**
+1. `mkdir -p <outputDir>`（`<config.outputDir>/<ctx.threadId>`）
+2. `const command = config.command ?? "office-agent"`
+3. `spawnCli(command, ["create", prompt, "-o", "output.docx"], { cwd: outputDir, timeoutMs })`
+4. 验证 `outputDir/output.docx` 存在
+5. 返回 `JSON.stringify({ mode: "generate", outputDocx, sourceDocx: null })`
+
+**编辑模式（`inputDocx ≠ null`）：**
+1. `mkdir -p <outputDir>`
+2. `copyFile(inputDocx, <outputDir>/original.docx)`
+3. `copyFile(inputDocx, <outputDir>/modified.docx)`
+4. `const command = config.command ?? "office-agent"`
+5. `spawnCli(command, ["edit", "modified.docx", prompt], { cwd: outputDir, timeoutMs })`
+6. 验证 `outputDir/modified.docx` 存在
+7. 返回 `JSON.stringify({ mode: "edit", outputDocx: modifiedPath, sourceDocx: originalPath })`
+
+### AdapterFn 实现（直接实现，不经过 runtime.extract）
+
+CLI 产出确定性 JSON，直接 `schema.parse(JSON.parse(raw))` 跳过 LLM extraction：
+
+```typescript
+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);
+      const raw = await runOfficeAgent(config, ctx.threadId, prompt, inputDocx);
+      const meta = schema.parse(JSON.parse(raw)) as T;
+      return { meta, childThread: null };
+    };
+}
+```
+
+`_systemPrompt` 为 writer 角色的 systemPrompt（空字符串），实际指令从 `ctx.start.content` 解析。
+
+### 配置
+
+```typescript
+type OfficeAgentConfig = {
+  outputDir: string;        // 输出根目录，runner 在此下按 threadId 建子目录
+  command: string | null;   // null → runner 内 resolve 为 "office-agent"
+  timeout: number | null;   // null → 不设超时；单位 ms
+};
+```
+
+### 错误处理
+
+```typescript
+if (!result.ok) {
+  const e = result.error;
+  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");
+  // "spawn_failed"
+  throw new Error(`office-agent: spawn failed: ${e.message}`);
+}
+if (!existsSync(expectedPath))
+  throw new Error(`office-agent: output file not found: ${expectedPath}`);
+```
+
+### packageDescriptor
+
+```typescript
+// src/package-descriptor.ts
+export const packageDescriptor: PackageDescriptor = {
+  name: "@uncaged/workflow-agent-office",
+  version: "0.1.0",
+  capabilities: ["office-agent-cli", "docx-generate", "docx-edit"],
+  configSchema: {
+    type: "object",
+    required: ["outputDir"],
+    properties: {
+      outputDir: { type: "string", description: "Root directory for workflow outputs." },
+      command:   { anyOf: [{ type: "string" }, { type: "null" }], description: "Path to office-agent CLI; null uses PATH." },
+      timeout:   { anyOf: [{ type: "number" }, { type: "null" }], description: "Timeout in ms; null means no limit." },
+    },
+    additionalProperties: false,
+  },
+};
+```
+
+### 包文件结构
+
+```
+packages/workflow-agent-office/
+  src/
+    types.ts                # OfficeAgentConfig, OfficeAgentOpt
+    runner.ts               # runOfficeAgent()（spawnCli 封装 + 文件验证）
+    agent.ts                # createOfficeAgent(): AdapterFn
+    package-descriptor.ts   # packageDescriptor
+    index.ts
+  __tests__/
+    runner.test.ts
+    agent.test.ts
+  package.json
+  tsconfig.json
+```
+
+### 依赖
+
+```json
+{
+  "@uncaged/workflow-protocol": "workspace:^",
+  "@uncaged/workflow-util": "workspace:^",
+  "@uncaged/workflow-util-agent": "workspace:^"
+}
+```
+
+---
+
+## 三、`workflow-agent-docx-diff`
+
+`differ` 角色专用执行器。从 `ctx.steps` 读取 `WriterMeta`，调用本地 `docx-diff` CLI。
+
+### docx-diff 退出码约定
+
+| 退出码 | 含义 | runner 处理 |
+|---|---|---|
+| 0 | 无差异 | 正常，验证 diffDocx 存在 |
+| 1 | 有差异 | 正常（显式处理为成功），验证 diffDocx 存在 |
+| 2+ | 错误 | throw |
+
+runner 收到 `SpawnCliError { kind: "non_zero_exit", exitCode: 1 }` 时视为成功，验证文件后继续；`exitCode >= 2` 才 throw。
+
+### 执行流程
+
+```
+1. 从 ctx.steps 找到 writer 步骤，读取 WriterMeta
+2. 验证 mode === "edit"（否则 throw）
+3. diffDocx = join(dirname(writer.outputDocx), "diff.docx")
+4. const command = config.command ?? "docx-diff"
+5. spawnCli(command,
+     [writer.sourceDocx, writer.outputDocx, "--output", "docx", "--out-file", diffDocx],
+     { cwd: null, timeoutMs: null })
+   exit 0 或 1 → 验证 diffDocx 存在
+   exit 2+ → throw
+6. 返回 JSON.stringify({ sourceDocx, modifiedDocx: writer.outputDocx, diffDocx })
+```
+
+### AdapterFn 实现（直接实现，不经过 runtime.extract）
+
+```typescript
+export function createDocxDiffAgent(config: DocxDiffAgentConfig = { command: null }): AdapterFn {
+  return <T>(_prompt: string, schema: z.ZodType<T>) =>
+    async (ctx: ThreadContext, _runtime: WorkflowRuntime): Promise<RoleResult<T>> => {
+      const writerStep = ctx.steps.find(s => s.role === "writer");
+      if (!writerStep) 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");
+      const raw = await runDocxDiff(config, writerMeta);
+      const meta = schema.parse(JSON.parse(raw)) as T;
+      return { meta, childThread: null };
+    };
+}
+```
+
+### 配置
+
+```typescript
+type DocxDiffAgentConfig = {
+  command: string | null;   // null → runner 内 resolve 为 "docx-diff"
+};
+```
+
+### packageDescriptor
+
+```typescript
+export const packageDescriptor: PackageDescriptor = {
+  name: "@uncaged/workflow-agent-docx-diff",
+  version: "0.1.0",
+  capabilities: ["docx-diff-cli", "docx-diff-report"],
+  configSchema: {
+    type: "object",
+    properties: {
+      command: { anyOf: [{ type: "string" }, { type: "null" }], description: "Path to docx-diff CLI; null uses PATH." },
+    },
+    additionalProperties: false,
+  },
+};
+```
+
+### 包文件结构
+
+```
+packages/workflow-agent-docx-diff/
+  src/
+    types.ts                # DocxDiffAgentConfig
+    runner.ts               # runDocxDiff()（exit 1 处理 + 文件验证）
+    agent.ts                # createDocxDiffAgent(): AdapterFn
+    package-descriptor.ts   # packageDescriptor
+    index.ts
+  __tests__/
+    runner.test.ts
+    agent.test.ts
+  package.json
+  tsconfig.json
+```
+
+### 依赖
+
+```json
+{
+  "@uncaged/workflow-protocol": "workspace:^",
+  "@uncaged/workflow-util-agent": "workspace:^",
+  "@uncaged/workflow-template-document": "workspace:^"
+}
+```
+
+---
+
+## 四、外部 bundle（外部 workspace 消费）
+
+```typescript
+import { createOfficeAgent } from "@uncaged/workflow-agent-office";
+import { createDocxDiffAgent } from "@uncaged/workflow-agent-docx-diff";
+import {
+  buildDocumentDescriptor,
+  documentWorkflowDefinition,
+} from "@uncaged/workflow-template-document";
+import { createWorkflow } from "@uncaged/workflow-runtime";
+import { getDefaultWorkflowStorageRoot } from "@uncaged/workflow-util";
+import { join } from "node:path";
+
+const outputDir = join(getDefaultWorkflowStorageRoot(), "outputs");
+
+export const descriptor = buildDocumentDescriptor();
+export const run = createWorkflow(documentWorkflowDefinition, {
+  adapter: createOfficeAgent({ outputDir, command: null, timeout: null }),
+  overrides: { differ: createDocxDiffAgent() },
+});
+```
+
+---
+
+## 不在范围内
+
+- 重试逻辑（失败直接 throw）
+- office-agent server 的启停管理（假设 server 已在运行）
+- docx-diff HTML/terminal 格式输出（仅 docx）
+- 跨机器执行（`inputDocx` 须为本机有效绝对路径）
@@ -0,0 +1,539 @@
+# `uwf` — Stateless Workflow CLI
+
+> 将 workflow 引擎降维为无状态单步 CLI。Workflow 是纯数据（CAS 节点），执行是单步原子操作，agent 是可插拔外部命令。
+
+---
+
+## 1. CLI Design
+
+### 1.1 命令总览
+
+```
+# thread 组
+uwf thread start <workflow> -p <prompt>     # 创建 thread，不执行
+uwf thread step  <thread-id> [--agent]      # 单步执行
+uwf thread show  <thread-id>                # thread-id → head 查询
+uwf thread list  [--all]                    # 列出活跃 threads（--all 含已归档）
+uwf thread kill  <thread-id>                # 终结 thread，归档
+
+# workflow 组
+uwf workflow put   <file.yaml>              # 注册 workflow（YAML → CAS）
+uwf workflow show  <workflow-id>            # 查看 workflow 定义
+uwf workflow list                           # 列出已注册 workflows
+```
+
+两组对称，各 3-4 个子命令。CAS 操作交给 `json-cas` CLI，不在 `uwf` 中重复。
+
+### 1.2 `uwf thread start`
+
+```bash
+uwf thread start <workflow> -p "Fix the login bug described in issue #42"
+```
+
+- `<workflow>` — workflow 名或 CAS hash
+- `-p` — 用户 prompt（必填）
+
+**输出（JSON to stdout）：**
+
+```jsonc
+{
+  "workflow": "4KNM2PXR3B1QW",   // workflow CAS hash (XXH64, 13-char Crockford Base32)
+  "thread": "01J7K9M2XNPQR5VWBCDF8G3H4T"      // ULID
+}
+```
+
+**做的事：**
+1. 解析 workflow（名字查 registry → CAS hash）
+2. 生成 thread ULID
+3. 写 StartNode 到 CAS
+4. 在 threads.yaml 中记录链头 → StartNode hash
+5. 输出 JSON
+
+### 1.3 `uwf thread step`
+
+```bash
+uwf thread step 01J7K9M2XNPQR5VWBCDF8G3H4T
+uwf thread step 01J7K9M2XNPQR5VWBCDF8G3H4T --agent "bunx uwf-cursor"
+```
+
+**输出（JSON to stdout）：**
+
+```jsonc
+{
+  "workflow": "4KNM2PXR3B1QW",
+  "thread": "01J7K9M2XNPQR5VWBCDF8G3H4T",
+  "head": "8FWKR3TN5V1QA",       // 新链头 StepNode 的 CAS hash
+  "done": false                    // true = moderator 返回 END，thread 已归档
+}
+```
+
+`done: true` 时 head 仍然有值（最后一个 StepNode），但 thread 已从 threads.yaml 移除。
+对已结束或不存在的 thread 调用 step 会报错（非 active thread）。
+
+详细信息通过 `uwf thread show <thread-id>` 或 `json-cas get <head>` 查看。
+
+**做的事：**
+1. 读链头 → 当前 StepNode（或 StartNode）
+2. 收集 thread 历史（遍历链）
+3. 调 moderator：评估 JSONata conditions → 得到下一个 role（或 END）
+4. 若 END → 归档 thread，输出最后链头，退出
+5. 确定 agent command（`--agent` override > config.yaml per-workflow/role > config.yaml defaultAgent）
+6. 调用：`<agent-cmd> <thread-id> <role>`，捕获 stdout 得到新 StepNode hash
+7. 更新链头指针
+8. 再次调 moderator（基于新 StepNode）判断 done
+9. 输出 JSON
+
+### 1.4 `uwf thread show`
+
+```bash
+uwf thread show 01J7K9M2XNPQR5VWBCDF8G3H4T
+```
+
+**输出（JSON to stdout）：**
+
+```jsonc
+{
+  "workflow": "4KNM2PXR3B1QW",
+  "thread": "01J7K9M2XNPQR5VWBCDF8G3H4T",
+  "head": "8FWKR3TN5V1QA",
+  "done": false
+}
+```
+
+纯 thread-id → head 查询。详细内容用 `json-cas get <head>` 或 `json-cas walk <head>` 查看。
+
+### 1.5 Agent CLI 协议
+
+每个 agent 是一个命令，接受 thread-id 和 role 两个参数：
+
+```bash
+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 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）
+- 所有配置从环境变量读（LLM model、API key、extractor config）
+- exit 0 = 成功，非 0 = 失败
+
+**stdout 输出：**
+
+```
+8FWKR3TN5V1QA
+```
+
+`uwf step` 拿到这个 hash 后更新链头指针、判断 done。
+
+---
+
+## 2. CAS 结构定义
+
+### 2.1 类型层级
+
+沿用 json-cas 的三层：bootstrap meta-schema → JSON Schema nodes → data nodes。
+
+下面所有 CAS 节点都遵循 `{ type: cas_ref, payload: T, timestamp: number }` 的标准格式。
+`cas_ref` 类型的字符串字段在 json-cas 中已内置支持，不需要额外的 `$ref` 包装。
+
+### 2.2 数据节点
+
+#### `Workflow`
+
+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>
+payload:
+  name: "solve-issue"
+  description: "End-to-end issue resolution"
+  roles:
+    planner:
+      description: "Creates implementation plan"
+      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"
+      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"
+      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"
+      expression: "$exists(steps[-1].output.needsClarification)"
+    notApproved:
+      description: "Reviewer rejected the implementation"
+      expression: "steps[-1].output.approved = false"
+  graph:
+    $START:
+      - role: "planner"
+        condition: null                  # 无条件（fallback）
+    planner:
+      - role: "developer"
+        condition: "needsClarification"
+      - role: "$END"
+        condition: null
+    developer:
+      - role: "reviewer"
+        condition: null
+    reviewer:
+      - role: "developer"
+        condition: "notApproved"
+      - role: "$END"
+        condition: null
+```
+
+- `roles` — 内联定义，每个 role 的 `meta` 是独立的 cas_ref（指向 json-cas 内置 JSON Schema 节点）
+- `conditions` — `Record<Name, JSONata>`，命名条件，方便画图描述
+- `graph` — `Record<Role | "$START", Transition[]>`，每个 Transition = `{ role, condition }`
+- `condition` 引用 conditions 中的 key，`null` = fallback
+- 按数组顺序求值，第一个匹配的 transition 胜出
+- 不含 agent binding — agent 配置在 `~/.uncaged/workflow/config.yaml` 中管理
+
+JSONata 表达式的求值上下文：
+
+```jsonc
+{
+  "start": {                          // StartNode 信息
+    "workflow": "4KNM2PXR3B1QW",
+    "prompt": "Fix the login bug..."
+  },
+  "steps": [                          // 所有已完成 steps，从旧到新
+    { "role": "planner", "output": { "phases": [...] }, "detail": "7BQST3VW9F2MA", "agent": "uwf-hermes" },
+    { "role": "developer", "output": { "filesChanged": ["src/auth.ts"], "summary": "Fixed redirect" }, "detail": "9KRVW3TN5F1QA", "agent": "uwf-cursor" },
+    { "role": "reviewer", "output": { "approved": false }, "detail": "2MXBG6PN4A8JR", "agent": "uwf-hermes" }
+  ]
+}
+```
+
+注：`output` 在上下文中会被自动展开为实际的 CAS 节点内容（而非 hash），方便 JSONata 表达式直接访问字段。
+
+#### `StartNode`（Thread 起点）
+
+```yaml
+type: <start-node-schema-hash>
+payload:
+  workflow: "4KNM2PXR3B1QW"        # cas_ref → Workflow
+  prompt: "Fix the login bug..."
+```
+
+- 没有 thread-id — thread-id 是索引层面的事，不进 CAS 内容
+- 没有 agent binding — 运行时从 config.yaml 解析
+
+#### `StepNode`（Thread 每一步）
+
+```yaml
+type: <step-node-schema-hash>
+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 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 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
+
+### 2.3 链式结构
+
+```
+threads.yaml: { "01J7K9M2XNPQR5VWBCDF8G3H4T": "8FWKR3TN5V1QA" }
+                                      │
+                                      ▼
+                              StepNode (step 3)
+                              ├── start ──→ StartNode
+                              │              ├── workflow → CAS(Workflow)
+                              │              └── prompt: "Fix..."
+                              ├── prev ──→ StepNode (step 2)
+                              │             ├── start ──→ (same StartNode)
+                              │             ├── prev ──→ StepNode (step 1)
+                              │             │             ├── start ──→ (same StartNode)
+                              │             │             ├── prev: null
+                              │             │             ├── role: "planner"
+                              │             │             └── ...
+                              │             ├── role: "developer"
+                              │             └── ...
+                              ├── role: "reviewer"
+                              ├── output → CAS({ approved: true })
+                              ├── detail → CAS(raw output | sub-workflow terminal node)
+                              └── agent: "uwf-hermes"
+```
+
+### 2.4 可变状态
+
+系统两个顶层 YAML 文件和一个 env 文件：
+
+```yaml
+# ~/.uncaged/workflow/config.yaml — 全局配置
+providers:
+  openai:
+    baseUrl: "https://api.openai.com/v1"
+    apiKeyEnv: "OPENAI_API_KEY"
+  anthropic:
+    baseUrl: "https://api.anthropic.com/v1"
+    apiKeyEnv: "ANTHROPIC_API_KEY"
+  openrouter:
+    baseUrl: "https://openrouter.ai/api/v1"
+    apiKeyEnv: "OPENROUTER_API_KEY"
+
+models:
+  sonnet:
+    provider: "openrouter"
+    name: "anthropic/claude-sonnet-4"
+  gpt4o-mini:
+    provider: "openai"
+    name: "gpt-4o-mini"
+
+agents:
+  hermes:
+    command: "uwf-hermes"
+    args: []
+  cursor:
+    command: "uwf-cursor"
+    args: []
+
+defaultAgent: "hermes"
+agentOverrides:
+  solve-issue:
+    developer: "cursor"
+
+defaultModel: "sonnet"
+modelOverrides:
+  extract: "gpt4o-mini"
+```
+
+```yaml
+# ~/.uncaged/workflow/threads.yaml — active thread 链头指针
+01J7K9M2XNPQR5VWBCDF8G3H4T: "8FWKR3TN5V1QA"
+01J8AB3QRMSTV6WKXZ2C4DF7GN: "3CNWT9KR6D2HV"
+```
+
+Thread 结束时从 threads.yaml 移除。可选：追加到 `history.jsonl` 做归档。
+
+```bash
+# ~/.uncaged/workflow/.env — 敏感信息（API keys）
+OPENAI_API_KEY=sk-...
+ANTHROPIC_API_KEY=sk-ant-...
+OPENROUTER_API_KEY=sk-or-...
+```
+
+- `config.yaml` — 非敏感配置（agent 命令、model 名、provider 名）
+- `.env` — 敏感信息（API keys），agent-kit 启动时自动加载
+- `threads.yaml` — 运行时状态
+
+---
+
+## 3. 包结构
+
+全新包，不复用现有 packages，避免命名冲突。CAS 直接依赖 `@uncaged/json-cas`。
+
+```
+packages/
+├── cli-workflow/              # @uncaged/cli-workflow — uwf CLI（thread/workflow 命令）
+├── workflow-moderator/        # @uncaged/workflow-moderator — JSONata moderator 引擎
+├── workflow-agent-kit/        # @uncaged/workflow-agent-kit — Agent CLI 框架（含 extractor）
+├── workflow-agent-hermes/     # @uncaged/workflow-agent-hermes — uwf-hermes CLI
+├── workflow-agent-cursor/ # @uncaged/workflow-agent-cursor — uwf-cursor CLI
+└── workflow-protocol/         # @uncaged/workflow-protocol — 共享类型定义
+```
+
+**外部依赖：**
+- `@uncaged/json-cas` — CAS 存储、hash、schema 校验
+- `@uncaged/json-cas-fs` — 文件系统 CAS 后端
+
+**现有包全部保留不动**，新旧并存，逐步迁移。
+
+---
+
+## 4. 关键数据类型
+
+JSONata 求值上下文本质上是 thread 链表的线性化表达。StepNode payload 和上下文中的 step 共享大量字段，提取为公共类型。
+
+### 4.1 公共类型
+
+```typescript
+/** CAS hash — XXH64, 13-char Crockford Base32 */
+type CasRef = string;
+
+/** Thread ID — ULID, 26-char Crockford Base32 */
+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 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）
+};
+```
+
+### 4.2 Workflow 定义
+
+```typescript
+type RoleDefinition = {
+  description: string;
+  goal: string;
+  capabilities: string[];
+  procedure: string;
+  output: string;
+  meta: CasRef;                      // cas_ref → json-cas 内置 JSON Schema 节点
+};
+
+type Transition = {
+  role: string;                      // 目标 role 名 或 "$END"
+  condition: string | null;          // 引用 conditions 中的 key，null = fallback
+};
+
+type ConditionDefinition = {
+  description: string;
+  expression: string;                           // JSONata expression
+};
+
+type WorkflowPayload = {
+  name: string;
+  description: string;
+  roles: Record<string, RoleDefinition>;
+  conditions: Record<string, ConditionDefinition>;
+  graph: Record<string, Transition[]>;          // Record<Role | "$START", Transition[]>
+};
+```
+
+### 4.3 Thread 节点
+
+```typescript
+type StartNodePayload = {
+  workflow: CasRef;                  // cas_ref → Workflow
+  prompt: string;
+};
+
+type StepNodePayload = StepRecord & {
+  start: CasRef;                     // cas_ref → StartNode（每个 step 都引用）
+  prev: CasRef | null;               // cas_ref → 前一个 StepNode，第一步为 null
+};
+```
+
+### 4.4 JSONata 求值上下文
+
+Thread 链表的线性化。`steps[n]` 的字段和 `StepRecord` 一致，但 `output` 被展开为实际内容。
+
+```typescript
+/** JSONata 上下文中的 step — output 被展开 */
+type StepContext = Omit<StepRecord, "output"> & {
+  output: unknown;                   // 展开后的 CAS 节点内容，非 hash
+};
+
+type ModeratorContext = {
+  start: StartNodePayload;
+  steps: StepContext[];              // 从旧到新
+};
+```
+
+### 4.5 CLI 输出
+
+```typescript
+/** uwf thread start */
+type StartOutput = {
+  workflow: CasRef;
+  thread: ThreadId;
+};
+
+/** uwf thread step / uwf thread show */
+type StepOutput = {
+  workflow: CasRef;
+  thread: ThreadId;
+  head: CasRef;
+  done: boolean;
+};
+
+/** uwf thread list */
+type ThreadListItem = {
+  thread: ThreadId;
+  workflow: CasRef;
+  head: CasRef;
+};
+```
+
+### 4.6 配置
+
+```typescript
+/** Alias types for config references */
+type AgentAlias = string;
+type ModelAlias = string;
+type ProviderAlias = string;
+type WorkflowName = string;
+type RoleName = string;
+type Scenario = string;              // e.g. "extract"
+
+type ProviderConfig = {
+  baseUrl: string;
+  apiKeyEnv: string;                 // env var name to read API key from
+};
+
+type ModelConfig = {
+  provider: ProviderAlias;
+  name: string;                      // e.g. "anthropic/claude-sonnet-4", "gpt-4o-mini"
+};
+
+type AgentConfig = {
+  command: string;
+  args: string[];
+};
+
+/** ~/.uncaged/workflow/config.yaml */
+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;
+};
+
+/** ~/.uncaged/workflow/threads.yaml */
+type ThreadsIndex = Record<ThreadId, CasRef>;
+//                         ^ thread-id  ^ head StepNode/StartNode hash
+```
+
+### 4.7 类型关系图
+
+```
+WorkflowConfig (config.yaml)
+ThreadsIndex (threads.yaml)          ← 唯二可变状态
+    │
+    │ thread-id → head hash
+    ▼
+StepNodePayload ──extends──→ StepRecord ←──maps to──→ StepContext
+    │                           │                          │
+    ├── start → StartNodePayload│                          │ (output 展开)
+    ├── prev → StepNodePayload  │                          │
+    │                           ├── role                   ├── role
+    │                           ├── output (CasRef)        ├── output (展开)
+    │                           ├── detail (CasRef)        ├── detail (CasRef)
+    │                           └── agent (string)         └── agent (string)
+    │
+    └── start.workflow → WorkflowPayload
+                             ├── roles: Record<name, RoleDefinition>
+                             ├── conditions: Record<name, JSONata>
+                             └── graph: Record<role, Transition[]>
+```
@@ -0,0 +1,41 @@
+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.
+    meta:
+      type: object
+      properties:
+        thesis:
+          type: string
+        keyPoints:
+          type: array
+          items:
+            type: string
+        caveats:
+          type: string
+      required: [thesis, keyPoints]
+conditions: {}
+graph:
+  $START:
+    - role: "analyst"
+      condition: null
+  analyst:
+    - role: "$END"
+      condition: null
@@ -0,0 +1,75 @@
+name: "solve-issue"
+description: "End-to-end issue resolution"
+roles:
+  planner:
+    description: "Creates implementation plan"
+    goal: "You are a planning agent. You analyze issues and create step-by-step plans."
+    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
+        steps:
+          type: array
+          items:
+            type: string
+      required: [plan, steps]
+  developer:
+    description: "Implements code changes"
+    goal: "You are a developer agent. You implement code changes according to plans."
+    capabilities:
+      - file-edit
+      - shell
+      - testing
+    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
+        summary:
+          type: string
+      required: [filesChanged, summary]
+  reviewer:
+    description: "Reviews code changes"
+    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."
+    meta:
+      type: object
+      properties:
+        approved:
+          type: boolean
+        comments:
+          type: string
+      required: [approved, comments]
+conditions:
+  notApproved:
+    description: "Reviewer rejected the implementation"
+    expression: "steps[-1].output.approved = false"
+graph:
+  $START:
+    - role: "planner"
+      condition: null
+  planner:
+    - role: "developer"
+      condition: null
+  developer:
+    - role: "reviewer"
+      condition: null
+  reviewer:
+    - role: "developer"
+      condition: "notApproved"
+    - role: "$END"
+      condition: null
@@ -0,0 +1,30 @@
+{
+  "name": "@uncaged/cli-workflow",
+  "version": "0.5.0-alpha.4",
+  "files": [
+    "src",
+    "dist",
+    "package.json"
+  ],
+  "type": "module",
+  "bin": {
+    "uncaged-workflow": "src/cli.ts"
+  },
+  "dependencies": {
+    "@uncaged/workflow-gateway": "workspace:^",
+    "@uncaged/workflow-protocol": "workspace:^",
+    "@uncaged/workflow-util": "workspace:^",
+    "@uncaged/workflow-cas": "workspace:^",
+    "@uncaged/workflow-execute": "workspace:^",
+    "@uncaged/workflow-register": "workspace:^",
+    "@uncaged/workflow-runtime": "workspace:^",
+    "hono": "^4.12.18",
+    "yaml": "^2.8.4"
+  },
+  "scripts": {
+    "test": "bun test"
+  },
+  "publishConfig": {
+    "access": "public"
+  }
+}
@@ -5,7 +5,6 @@ import { formatCliUsage as formatCliUsageWithGroups } from "./cli-usage.js";
 import { createCasDispatcher } from "./commands/cas/index.js";
 import { dispatchConnect } from "./commands/connect/index.js";
 import { createInitDispatcher } from "./commands/init/index.js";
-import { createJsonCasDispatcher } from "./commands/json-cas/index.js";
 import { dispatchSetup } from "./commands/setup/index.js";
 import { createThreadDispatcher, dispatchLive, dispatchRun } from "./commands/thread/index.js";
 import { createWorkflowDispatcher } from "./commands/workflow/index.js";
@@ -44,7 +43,6 @@ const dispatchWorkflow = createWorkflowDispatcher({ dispatchGroup });
 const dispatchThread = createThreadDispatcher({ dispatchGroup });
 const dispatchCas = createCasDispatcher({ dispatchGroup });
 const dispatchInit = createInitDispatcher({ dispatchGroup });
-const dispatchJsonCas = createJsonCasDispatcher({ dispatchGroup });

 async function showSkillDocOrIndex(topic: string | undefined): Promise<number> {
  if (topic === undefined) {
@@ -74,7 +72,6 @@ const COMMAND_TABLE: Record<string, DispatchFn> = {
  run: dispatchRun,
  live: dispatchLive,
  connect: dispatchConnect,
-  "json-cas": dispatchJsonCas,
 };

 export async function runCli(storageRoot: string, argv: string[]): Promise<number> {
@@ -2,7 +2,6 @@ import type { CommandGroup } from "./cli-command-types.js";
 import { setCommandGroupsForUsage } from "./cli-usage-context.js";
 import { CAS_SUBCOMMAND_TABLE } from "./commands/cas/index.js";
 import { INIT_SUBCOMMAND_TABLE } from "./commands/init/index.js";
-import { JSON_CAS_SUBCOMMAND_TABLE } from "./commands/json-cas/index.js";
 import { THREAD_SUBCOMMAND_TABLE } from "./commands/thread/index.js";
 import { WORKFLOW_SUBCOMMAND_TABLE } from "./commands/workflow/index.js";

@@ -53,14 +52,6 @@ export function getCommandRegistry(): ReadonlyArray<CommandGroup> {
      name: "setup",
      commands: [...SETUP_USAGE_COMMANDS],
    },
-    {
-      name: "json-cas",
-      commands: Object.entries(JSON_CAS_SUBCOMMAND_TABLE).map(([name, e]) => ({
-        name,
-        args: e.args,
-        description: e.description,
-      })),
-    },
  ];
 }

@@ -13,7 +13,6 @@ const USAGE_SECTION_BY_GROUP: Record<string, string> = {
  cas: "Content-addressable storage:",
  init: "Development:",
  setup: "Configuration:",
-  "json-cas": "JSON-CAS engine:",
 };

 export function formatUsageCommandLines(
@@ -0,0 +1,9 @@
+#!/usr/bin/env bun
+
+import { runCli } from "./cli-dispatch.js";
+import { resolveWorkflowStorageRoot } from "./storage-env.js";
+
+const argv = process.argv.slice(2);
+const storageRoot = resolveWorkflowStorageRoot();
+const code = await runCli(storageRoot, argv);
+process.exit(code);
--- a/Show More
+++ b/Show More