Store the fully assembled prompt sent to each agent in CAS as a text
node, referenced from StepNodePayload.assembledPrompt. This enables
exact reproduction of what the agent received for debugging hallucinations.
Changes:
- workflow-protocol: StepRecord + STEP_NODE_SCHEMA add assembledPrompt field
- workflow-util-agent: AgentRunResult includes assembledPrompt, run.ts stores it
- workflow-util-agent: schemas register TEXT_SCHEMA for prompt storage
- workflow-agent-claude-code: return assembled prompt from buildClaudeCodePrompt
- workflow-agent-hermes: return assembled prompt from buildHermesPrompt
- workflow-agent-builtin: return empty prompt (no prompt assembly)
- cli-workflow: step read --prompt renders the stored prompt
- All test fixtures updated for new field
Legacy steps without assembledPrompt show 'Prompt not recorded' message.
小橘 🍊
When mustache variables in edge prompts resolve to empty strings (because
upstream output lacks the fields), the engine now returns a Result.error
instead of passing an empty --prompt to the agent.
- evaluate.ts: check rendered prompt is non-empty after mustache.render()
- run.ts: improve parseArgv error message for empty --prompt
- Export parseArgv for testability
- Add 7 tests covering all cases from the spec
Implement thread-level and edge-level working directory management:
- Thread-level cwd (required, defaults to process.cwd())
- Captured at uwf thread start time
- Stored in StartNodePayload
- Inherited by all steps unless overridden
- Edge-level location (optional, supports mustache templates)
- New location: string | null field on Target type
- Resolved by moderator using previous step's output
- Example: location: "{{{repoPath}}}"
- Step audit trail
- Each StepNodePayload records actual cwd where agent executed
Changes:
- workflow-protocol: Add cwd to StartNodePayload & StepRecord, location to Target
- cli-workflow: Thread start captures cwd, moderator resolves location, step execution uses resolved cwd
- workflow-util-agent: Expose cwd in agent context
Tests:
- Protocol type tests (3 scenarios)
- Moderator location resolution tests (5 scenarios)
- Thread-location integration tests (3 scenarios)
All tests pass. Build successful. Backward compatible.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
xiaoju