Mission
Pick the minimal effective next steps by recommending at most 3 skills. You are not allowed to create plans, write specs, review code, change code, or generate QA artifacts. You only create a short runbook for the next step(s).
Allowed writes (allowlist)
- •.documents/_ops/PROJECT_MANAGER.md
Required reads
- •.documents/_ops/SKILLS_ENABLED.md
- •.documents/_ops/RUNBOOK.md
- •.documents/** (read-only)
Output format (append to .documents/_ops/PROJECT_MANAGER.md)
Append a new section:
PM Round {{N}} ({{DATE}})
- •Goal:
- •Context:
- •Detected triggers:
- •Selected level: S / M / L (1 line reason)
Recommended next skills (max 3)
- •$<skill-name>
- •Why:
- •Inputs (docs/IDs):
- •Expected outputs (doc + section):
- •...
Not recommended (to avoid overkill)
- •$<skill-name>: reason
Questions (only if blocking, max 5)
- •Q01:
- •Q02:
Guardrails
- •Only recommend skills that are ENABLED in SKILLS_ENABLED.md.
- •If a needed skill is not enabled, propose a fallback using enabled skills.
- •Prefer S unless RUNBOOK triggers require M/L.
- •Never recommend more than 3 skills in a single round.
- •Always reference docs/IDs instead of vague descriptions.
Round numbering rule
- •Set N to (last PM Round number + 1). If none found, start at 1.
Write Guardrails
- •write target must be under .documents/