메뉴
This skill should be used when the user asks for "LLM council", "agents arguing", "decision debate", "model council", "advisory board", "council of agents", or wants multiple LLMs to debate an architecture, product, strategy, or technical decision.
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
| name | subagent-council |
| description | This skill should be used when the user asks for "LLM council", "agents arguing", "decision debate", "model council", "advisory board", "council of agents", or wants multiple LLMs to debate an architecture, product, strategy, or technical decision. |
| version | 0.1.0 |
Run advisory deliberation through codex-subagent. Use this skill for decisions, tradeoffs, strategy, architecture, product framing, and ambiguous technical choices. A council is not a vote. It is an issue ledger, adversarial positions, rebuttal, and synthesis.
context-pack/v1 with mode: council.role-card/v1 for each role.subagent-result/v1 structured output for every council role.For context and role details, read ../references/context-and-roles.md when needed.
architect: boundaries, ownership, governance, reversibility.critic: failure modes, hidden assumptions, security/safety concerns.technician: feasibility, implementation constraints, timeline realism.pragmatist: scope discipline, MVP cut lines, cost of delay.engineer: concrete build path, edge cases, testability.codex-subagent context build \
--subject "<decision>" \
--mode council \
--goal "Map tradeoffs and recommend a direction" \
--rule AGENTS.md \
--file decision.md \
--out context.json
codex-subagent panel run pi \
--context-pack context.json \
--role architect \
--role critic \
--role technician \
--role pragmatist \
--role engineer
codex-subagent panel results <panel-id> --structured
Return:
assessed, partial, or not_assessedThis skill should be used when the user asks for "autoresearch", "research loop", "experiment loop", "agentic research", "benchmark variants", "try approaches overnight", or wants subagents to run hypotheses, experiments, metrics, and logs.
This skill should be used when the user asks for "development with subagents", "parallel coding agents", "use pi/opencode/gsd2 workers", "subagent implementation", "delegate coding work", or wants explorer/planner/worker/tester/reviewer agents coordinated from Codex.
This skill should be used when the user asks for "multi-plane review", "review with different LLMs", "pi/opencode review panel", "review code/spec/idea through subagents", or wants independent reviewers for code, product, UX, evidence, security, or requirements.