菜单
Run the council review loop on a PR. Spawns all 5 reviewers (Architect, Minimalist, Craftsperson, Steward, Marketing Guru) on every PR, collects verdicts, fixes feedback, and repeats until all approve or the Architect calls it. Use after opening a PR or when AGENTS.md says to run council review.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Run the product council review loop on an opportunity or JTBD spec. Spawns all 7 product reviewers (Opportunist, Advocate, Economist, Skeptic, Narrator, Copywriter, Channel Scout) in parallel, collects verdicts. The CPO (see agents/product-council/cpo.md) synthesizes and breaks ties. Repeats until all approve or the CPO calls it. Use when evaluating opportunities, reviewing JTBD specs, or conducting outcome reviews.
Run the Opportunist's discovery workflow to find new product opportunities. Scans developer communities, competitive landscape, market signals, and existing user feedback. Produces Opportunity Briefs for anything with ≥2 independent signals. Use when you need new opportunities to evaluate, or when the CPO has an open goal slot to fill.
| name | council-review |
| description | Run the council review loop on a PR. Spawns all 5 reviewers (Architect, Minimalist, Craftsperson, Steward, Marketing Guru) on every PR, collects verdicts, fixes feedback, and repeats until all approve or the Architect calls it. Use after opening a PR or when AGENTS.md says to run council review. |
Spawns all five council members, collects verdicts, fixes feedback, repeats until unanimous or the Architect calls it. Uses Ralph Wiggum for iteration control — write .ralph/<name>.md before calling ralph_start.
See agents/implementor/AGENTS.md § "Council Review" under The Workflow for member details, persona files in agents/council/, and the spawn template.
All five always run: Architect, Minimalist, Craftsperson, Steward, and Marketing Guru. The Marketing Guru reviews every PR to ensure documentation, changelogs, and public-facing materials stay current with what's shipping.
Use subagent with parallel tasks array — one per reviewer. Each reviewer must:
DATABASE_URL=postgresql://remarq:remarq@localhost:5433/remarq npm testgh pr comment <number> --body "## <emoji> <Name> — Round N Review\n\n...verdict"If tests fail, fix them before spawning reviewers. Don't waste reviewer time on broken code.
Check PR comments for Round N verdicts from all spawned reviewers.
If round 5+, add to the Architect's prompt: "This is round 5+. If remaining objections are nit-picks, call ship it."