메뉴
Plan-phase Review Gate debate — pi-messenger threads, lane YAML, bus tools for parent orchestrator.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Commit staged git changes with configurable message format and Co-authored-by trailer from .pi/auto-commit.json. Use when the user asks to commit, when /harness-auto or harness run completion requires a commit, or before any git commit — never use raw git commit -m.
Compile task-specific harness context using context-mode and graphify. Use for architecture questions, large codebase context before harness-plan or harness-run. NEVER use lean-ctx on harness paths.
Structured user decisions via ask_user for harness setup, planning, and governance forks. Use with /harness-setup, /harness-plan, harness-auto plan phase, and when agents emit human_required.
Deprecated — use harness-review skill and /harness-review for the full post-run gate. This file remains as a pointer for older prompts.
Enforce harness governance phases, policy gates, budgets, and promotion rules. Use during execute, evaluate, merge, or when interpreting policy-gate / budget-guard messages.
Bootstrap ls-lint filename rules for harness projects — seed naming.manifest.json, generate merge-safe .ls-lint.yml, and document bootstrap vs --force sync. Use during /harness-setup, when adding ls-lint to a repo, or when .ls-lint.yml is missing or out of date.
| name | harness-debate-plan |
| description | Plan-phase Review Gate debate — pi-messenger threads, lane YAML, bus tools for parent orchestrator. |
Review Gate RACI: parent is chair; lane agents provide structured evidence in sequence.
Use when running Phase 5 of /harness-plan — Fagan-style structured inspection per focus (spec | wbs | schedule | quality). Parent is chair; within-round dialogue (claims → rebuttals → clarifications → counters → integrate).
| Agent | Role |
|---|---|
hypothesis-validator | Blind verifier (R1 only) |
plan-evaluator | Inspector (checklist) |
plan-adversary | Red team |
sprint-contract-auditor | DoD auditor (quality or round ≥4) |
review-integrator | Recorder / integration PM |
Do not add agents for fast profile — reduce focuses/rounds only.
| Profile | Mode | Focuses | When |
|---|---|---|---|
full | threaded | all four | High risk, fork, open questions |
standard | threaded | all four | Default med risk |
light | threaded | spec, quality | Low risk, high-confidence research |
fast | consolidated | spec, quality (one round) | Clear stack, no open questions; escalate to threaded on blockers |
Eligibility: harness_plan_debate_eligibility then harness_debate_open({ debate_profile, required_focuses }).
harness_debate_open({})
plan-<run_id> (tool normalizes wrong ids)..pi/harness/runs/<run_id>/debate-messenger/.Budget caps vary by profile (see plan-debate-eligibility.ts); standard plan profile uses min_focus_rounds=4, debate_global_cap=80000.
Call harness_debate_focus_coverage until all required focuses (from eligibility) appear in submitted review rounds and last review_gate_ready: true.
hypothesis-validator (blind verifier) before inspector.plan-evaluator (inspector) → lane + messenger claim.harness_messenger_read_round → plan-adversary (red team) → rebuttal.unresolved_claim_ids and exchange_count < max for profile.sprint-contract-auditor (DoD) when focus is quality or round ≥ 4.review-integrator (recorder) → harness_debate_submit_round.One subagent per subagent call — never batch debate lanes.
Lane YAML + messenger messages auto-apply on subagent complete (harness-debate-next-step). Fallback: harness_debate_apply_lane.
Resume: harness_debate_round_status({ round_index: N }) → run listed next_tool.
| kind | from | when |
|---|---|---|
| claim | PlanEvaluatorAgent | after evaluator lane |
| rebuttal | PlanAdversaryAgent | in_reply_to claim ids |
| clarification | PlanEvaluatorAgent | addresses open claims |
| counter | PlanAdversaryAgent | final pass; concede or dispute |
| integrate | ReviewIntegratorAgent | on submit_round |
harness_debate_consensus when focus coverage complete. approve_plan is hard-gated on lanes, messenger dialogue completeness, bus rounds, consensus not block.
Do not approve_plan on policy_decision: block. On human_required → ask_user first.
Rubrics: use the focus-specific checklist ids passed by the parent for the active round.