Reusable skills for AI coding agents — Claude Code, Codex, Gemini CLI, and Copilot CLI.
Send an implementation plan to another AI agent for iterative review. The host agent revises the plan based on reviewer feedback and re-submits until the reviewer approves. Supports all four major coding CLIs as reviewers with session resume, model/effort selection, and structured output parsing.
Features:
- Cross-agent review loop (max 5 rounds) with verdict parsing
- Provider-specific adapters: Codex (stdin + JSONL), Gemini (JSON), Claude Code (JSON), Copilot (JSONL)
- Session resume for multi-round context continuity
- Process tree kill on timeout/cancellation
- Automatic model and effort detection from reviewer output
Usage: /peer-plan-review <codex|gemini|claude|copilot> [model] [effort]
Copy the skill directory into your Claude Code skills folder:
cp -r peer-plan-review ~/.claude/skills/Or symlink it:
ln -s "$(pwd)/peer-plan-review" ~/.claude/skills/peer-plan-reviewThen add the skill to your Claude Code settings (.claude/settings.json):
{
"skills": ["~/.claude/skills/peer-plan-review/SKILL.md"]
}The agents/openai.yaml file provides the Codex agent configuration.
MIT