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。
直接命令不会经过审查 Prompt;运行前请先检查来源。
npx skills add https://github.com/katarmal-ram/codemoot --skill build命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地?可下载 SkillsMP 当前能够提供的文件。
Bidirectional AI slop scanner — Claude + GPT independently analyze, then debate disagreements.
Get an independent GPT review via Codex CLI. Use when you want a second opinion on code, plans, or architecture from a different AI model.
Real Claude vs GPT multi-round debate. Use when you need a second opinion, want to debate architecture decisions, or evaluate competing approaches with multi-model collaboration.
正在显示 SKILL.md
基于 SOC 职业分类
| name | build |
| description | Autonomous build loop — debate, plan, implement, review, fix — all in one session with GPT review. |
| user-invocable | true |
/build <task description>
Full pipeline: debate approach with GPT → user approval → implement → GPT review → fix → re-review until approved. SQLite tracking throughout.
codemoot build start "TASK"Use /debate protocol. Loop until GPT says STANCE: SUPPORT.
Before user approval, validate the plan with GPT:
codemoot plan review /path/to/plan.md --build BUILD_ID
GPT reviews the plan against actual codebase, returns ISSUE/SUGGEST lines. Fix HIGH issues before presenting to user.
Present agreed plan. Wait for explicit approval via AskUserQuestion.
codemoot build event BUILD_ID plan_approved
codemoot debate complete DEBATE_ID
Write code. Run tests: pnpm run test
Never send broken code to review.
codemoot build event BUILD_ID impl_completed
codemoot build review BUILD_ID
Parse verdict: approved → Phase 4.5, needs_revision → Phase 4
Fix every CRITICAL and BUG. Run tests. Back to Phase 3.
codemoot build event BUILD_ID fix_completed
Compare deliverables against original request. Every requirement must be met.
codemoot build status BUILD_ID
Present summary with metrics, requirements checklist, GPT verdict.