review-loop — Claude Code skill: worker-reviewer cycle
A skill that runs a worker-reviewer loop: a separate critic subagent scores the work 1-10 and hands back notes, and the agent revises until the gate is met.
npx skills add 2dmurali/review-loop-skill --skill review-loop -a claude-codeWhat review-loop does
You finish the task the way you normally would. Then the skill hands the result to a critic subagent, which sees only the work product — files and diffs, not your reasoning — and returns a score from 1 to 10 with pointed feedback. The agent revises against those notes and the cycle repeats.
Defaults are 2 loops minimum, 4 maximum, quality gate 8/10. All three can be changed in the request itself: "quality gate 9", "max 5 loops", "reasoning reviewer". The reviewer has to be at least as capable as the worker, otherwise subtle regressions slip past it. If the score stalls across two rounds on the same feedback, the worker model is the bottleneck — swap it or split the task.
The skill assumes the platform can spawn subagents. Where it cannot, the review step is faked by hand: open a fresh chat, paste the work product alone, ask for a score and specific notes. Quick edits and tasks gated by tests are a poor fit.
What it helps with
- turn a rough spec draft into something readable
- get a second agent's read on code instead of grading your own work
- set a quality gate and loop count per task
- harden an implementation where reliability beats speed
- catch regressions across a multi-file change
Example prompts for the agent
- Implement report export, use review-loop
- Run review-loop on the file I just wrote
- Polish this spec, quality gate 9, max 5 loops
How to install review-loop
1Skills CLI (Vercel)
npx skills add 2dmurali/review-loop-skill --skill review-loop -a claude-code2Manually: put the files in place
~/.claude/skills/review-loop/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/review-loop/
1Skills CLI (Vercel)
npx skills add 2dmurali/review-loop-skill --skill review-loop -a codex2Manually: put the files in place
.agents/skills/review-loop/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add 2dmurali/review-loop-skill --skill review-loop -a cursor2Manually: put the files in place
.cursor/skills/review-loop/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install 2dmurali/review-loop-skill review-loop2Manually: put the files in place
.github/skills/review-loop/SKILL.md1Skills CLI (Vercel)
npx skills add 2dmurali/review-loop-skill --skill review-loop -a gemini-cli2Platform CLI
gemini skills install https://github.com/2dmurali/review-loop-skill --path review-loopManually: put the files in place
.devin/skills/review-loop/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add 2dmurali/review-loop-skill --skill review-loop -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about review-loop
How do I install review-loop in Claude Code?
Run npx skills add 2dmurali/review-loop-skill --skill review-loop -a claude-code. The skill lands at ~/.claude/skills/review-loop/SKILL.md.
What does review-loop do?
It runs a worker-reviewer cycle inside one session: a critic subagent scores the output 1-10 with actionable feedback, the agent revises, and the loop continues until the quality gate is reached.
What does review-loop need to work?
A platform that can spawn subagents. Without that, the review step is done manually: a fresh chat, the work product only, and a request to score it with specific feedback.
