councillisted

# Council Code Review Skill This skill runs the same code review prompt against multiple AI agents in parallel, then synthesizes their findings into a single comprehensive report. It supports **Claude Code**, **Codex CLI**, **OpenCode**, and **Cursor** as agent backends -- and you can mix them freely. ## When to Use Activate this skill when the user asks to: - "Review my code" - "Run a code review" - "Review the staged changes" - "Do a council review" - "Get multiple perspectives on this code" ## CRITICAL: Target Directory **You must pass the USER'S PROJECT DIRECTORY as an argument to the script.** The user's project directory is where they started their agent session - NOT this skill's directory. Look for the git repository path in the conversation context (e.g., `/Users/.../git/jupyter_server`). ## Workflow ### Step 1: Identify the Target Repository Determine the user's project directory from the conversation context. This is typically shown at the start of the session or can be found by checking where `AGENTS.md`, `CLAUDE.md`, or similar agent guidance is located. It is NOT `/Users/.../skills/council/`. ### Step 2: Run Parallel Reviews Run this skill's review script and **pass the user's project directory as an argument**: ```bash ~/.agents/skills/council/scripts/run-reviews.sh /path/to/users/project ``` **IMPORTANT**: Always pass the full path to the user's project as the first argument. #### Passing review instructions (target branch, focus areas, etc.) If