phase-plan

Phase Planning Workflow

YOU ARE A PURE ORCHESTRATOR. Spawn phase-planner, parse its JSON response, track workflow state, and register the generated artifact. Do not write or edit planning artifacts yourself.

Command Contract: /phase-plan <source> [--afk]

§CTX

Use the pre-resolved projectRoot, kbRoot, and workRoot values from the generated Workflow Bootstrap section. Do not hardcode .rp1/work/ or .rp1/context/ paths.

SOURCE is the user-provided planning source path or identifier. It must resolve to either:

• a PRD under {workRoot}/prds/*.md
• an oversized requirements artifact under {workRoot}/features/*/requirements.md

UPDATE_CONTEXT is optional revision guidance for refreshing an existing phase plan.

This workflow is single-pass. It does not emit waiting_for_user for source ambiguity. If SOURCE is ambiguous, the planner must return terminal status="error" JSON so the user can rerun /phase-plan with an explicit source path.

STATE-MACHINE

stateDiagram-v2
    [*] --> analyze
    analyze --> publish : plan_ready
    publish --> completed : published
    analyze --> failed : error
    publish --> failed : error
    completed --> [*]
    failed --> [*]

First emit (entering analyze) includes --name:

rp1 agent-tools emit \
  --workflow phase-plan \
  --type status_change \
  --run-id {RUN_ID} \
  --step analyze \
  --name "Phase Plan: {source summary}" \
  --data '{"status": "running"}'

Subsequent transitions omit --name:

rp1 agent-tools emit \
  --workflow phase-plan \
  --type status_change \
  --run-id {RUN_ID} \
  --step {STATE} \
  --data '{"status": "running"}'

Terminal states:

rp1 agent-tools emit \
  --workflow phase-plan \
  --type status_change \
  --run-id {RUN_ID} \
  --step {STATE} \
  --data '{"status": "completed"}'

On failure:

rp1 agent-tools emit \
  --workflow phase-plan \
  --type status_change \
  --run-id {RUN_ID} \
  --step failed \
  --data '{"status": "failed"}'