do-and-judge — Claude Code skill: what it does, setup
This skill sends a task to an implementation sub-agent, checks the result with an independent judge, and retries with feedback when needed.
npx skills add NeoLabHQ/context-engineering-kit --skill do-and-judge -a claude-codeWhat do-and-judge does
do-and-judge separates implementation from verification. The orchestrator sends the task to an implementation sub-agent, asks a second sub-agent to prepare evaluation criteria in parallel, then gives the result to an independent judge. The judge returns a score and issues to address.
When verification fails, the skill starts another implementation attempt with the judge's feedback. You can select a model with --model or require a passing score of at least 4.0 with --strict. Check the retry limit before using it: after 3 failed attempts, the workflow requires a user decision.
What it helps with
- delegate a task to a separate implementation sub-agent
- generate task-specific evaluation criteria
- check a result with an independent judge
- retry work using the judge's feedback
Example prompts for the agent
/do-and-judge Refactor the UserService class to use dependency injection/do-and-judge Review error handling in the API --strict/do-and-judge Add tests for the new function --model sonnet
How to install do-and-judge
1Skills CLI (Vercel)
npx skills add NeoLabHQ/context-engineering-kit --skill do-and-judge -a claude-code2Manually: put the files in place
~/.claude/skills/do-and-judge/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/do-and-judge/
1Skills CLI (Vercel)
npx skills add NeoLabHQ/context-engineering-kit --skill do-and-judge -a codex2Manually: put the files in place
.agents/skills/do-and-judge/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add NeoLabHQ/context-engineering-kit --skill do-and-judge -a cursor2Manually: put the files in place
.cursor/skills/do-and-judge/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install NeoLabHQ/context-engineering-kit do-and-judge2Manually: put the files in place
.github/skills/do-and-judge/SKILL.md1Skills CLI (Vercel)
npx skills add NeoLabHQ/context-engineering-kit --skill do-and-judge -a gemini-cli2Platform CLI
gemini skills install https://github.com/NeoLabHQ/context-engineering-kit --path skills/do-and-judgeManually: put the files in place
.devin/skills/do-and-judge/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add NeoLabHQ/context-engineering-kit --skill do-and-judge -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 do-and-judge
How do I install do-and-judge in Claude Code?
Run npx skills add NeoLabHQ/context-engineering-kit --skill do-and-judge -a claude-code, then run the same command again: npx skills add NeoLabHQ/context-engineering-kit --skill do-and-judge -a claude-code. Check ~/.claude/skills/do-and-judge/SKILL.md afterward.
What does do-and-judge do?
The skill delegates a task to an implementation sub-agent, creates evaluation criteria, and sends the result to an independent judge. A failed check starts another attempt with specific feedback.
Which platforms support do-and-judge?
The skill supports Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, and OpenCode.
