| name | consensus-planning |
| description | Build an implementation-ready plan by iterating with peer-review until concerns converge |
What I do
- Turn discussion context into an implementation-ready plan.
- Run an explicit review loop with the
peer-review agent to challenge assumptions and tradeoffs.
- Keep only unresolved high-impact concerns in scope each round.
When to use me
- The user needs a robust plan for non-trivial implementation work.
- There are architecture or sequencing decisions with meaningful risk.
- You want convergence through critique, not a one-shot plan.
Workflow
- Draft an initial plan from current conversation context.
- Run a review round with the
peer-review agent.
- Revise the plan to address feedback and record what changed.
- Repeat steps 2-3 until convergence or round cap.
Round cap:
Convergence criteria (stop when any condition is true):
- The
peer-review agent explicitly indicates convergence.
- No new high-priority concerns appear for 2 consecutive rounds.
- The process reaches 6 rounds.
Per-round rules
- Use the
peer-review agent each round; do not rely on self-critique alone.
- Ask peer-review to challenge at least one already accepted decision.
- Tag concerns as
new, reframed, or resolved.
- Prioritize high-impact risks and sequencing problems first.
- Do not repeat resolved points unless new evidence appears.
- If unknowns remain, choose best-effort defaults and list targeted follow-up questions.
Output format
- Goal
- Decisions taken
- Ordered action plan
- Risks and mitigations
- Verification checklist
- Consensus summary (rounds used, stop reason, confidence 1-5)
- Open questions (if any)
Keep the final output concise, concrete, and executable.