HN 표시: Cook을 사용하여 작업->검토 루프에서 Claude를 자동화합니다.

A simple CLI for orchestrating Claude Code, Codex, and OpenCode. # review loop cook "Implement dark mode" review # 3 passes cook "Implement dark mode" x3 # race 3, pick best cook "Implement dark mode" v3 "least code" # two approaches, pick one cook "Auth with JWT" vs "Auth with sessions" pick "best security" # task list cook "Work on next task in plan.md" review ralph 5 "DONE if all tasks complete, else NEXT" Two ways to get it: Skill (no CLI) — your agent orchestrates subagents directly: mkdir -p .claude/skills/cook && curl -fsSL https://raw.githubusercontent.com/rjcorwin/cook/main/no-code/SKILL.md -o .claude/skills/cook/SKILL.md CLI — standalone tool for terminal or CI: npm install -g @let-it-cook/cli mkdir -p .claude/skills && cp -r $(npm root -g)/@let-it-cook/cli/skill .claude/skills/cook Requires Node.js 20+ and Claude Code, Codex, or OpenCode. By default, Docker sandbox restricts outbound traffic to only what each agent requires. Configure in .cook/config.json : { "docker": { "network": { "mode": "restricted", "allowedHosts": ["api.mycompany.com"] } } } mode | Behavior | |---|---| "restricted" (default) | Blocks all outbound traffic except DNS and HTTPS to a whitelist of required agent hosts | "unrestricted" | No network restrictions | allowedHosts adds extra hosts to the whitelist when mode is "restricted" . cook init # Set up COOK.md, config, and Dockerfile cook rebuild # Rebuild the sandbox Docker image cook doctor # Check Docker + auth readiness cook shell / sandbox # Interactive shell in Docker sandbox cook shell # Run a command in the sandbox cook shell --unrestricted # Shell with unrestricted networking Work — a prompt string. One agent call. Loop operators wrap work with iteration: xN — N sequential passesreview — review → gate → iterate loop (up to 3 by default)ralph N "" — repeat until a gate says DONE Composition operators run parallel branches then resolve: vN — N identical branches, pick the bestvs — two different branches, pick the bestpick ,merge ,compare — resolvers Operators compose left to right. Each wraps everything to its left. cook "work" x3 review # 3 passes, then review loop cook "work" review x3 # review loop, repeated 3 times cook "work" review v3 # race 3, each with a review loop cook "Add dark mode" review # default prompts, up to 3 iterations cook "Add dark mode" review 5 # up to 5 iterations cook "Add dark mode" "Review for accessibility" "DONE if WCAG AA, else ITERATE" cook "Add dark mode" review --work-agent codex --review-agent claude --review-model opus When an agent hits a token quota or rate limit, cook automatically waits and retries instead of bailing. A countdown is shown in the TUI. Enabled by default. cook "Build the feature" review --no-wait # disable: fail immediately on rate limit Configure in .cook/config.json : { "retry": { "enabled": true, "pollIntervalMinutes": 5, "maxWaitMinutes": 360 } } cook init Creates COOK.md (project instructions + prompt template) and .cook/config.json : { "agent": "claude", "sandbox": "agent", "steps": { "work": { "agent": "codex", "model": "gpt-5-codex" }, "review": { "agent": "claude", "model": "opus" } }, "env": ["CLAUDE_CODE_OAUTH_TOKEN"] }