Skip to main content
affpapa.org
RU EN
skill CX Codex

judge — Claude Code skill: evaluation and installation

The skill runs a two-stage review of work from the current conversation: one sub-agent sets the criteria, another scores the result with evidence.

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓ 
npx skills add NeoLabHQ/context-engineering-kit --skill judge -a claude-code
Open on GitHub

What judge does

judge extracts the task, produced result, involved files, and acceptance criteria from the current conversation. It then runs two stages: a meta-judge prepares a tailored review rubric, while a judge sub-agent applies it in a separate context. This keeps the assessment apart from the session state built up earlier.

The review uses structured scores, checklists, and evidence tied to concrete work locations, including files and lines. The skill returns findings without changing the project automatically. You still need to handle any fixes yourself.

What it helps with

  • review code after an implementation task
  • create criteria for checking documentation
  • verify a result against files and line references
  • get a structured sub-agent report
  • spot gaps between a score and its reasoning

Example prompts for the agent

  • Evaluate the result of the current coding task and cite evidence from the files.
  • Review the documentation: create the criteria first, then run a separate judgment.
  • Produce a quality report for the changed files without applying automatic fixes.

How to install judge

1Skills CLI (Vercel)

npx skills add NeoLabHQ/context-engineering-kit --skill judge -a claude-code

2Manually: put the files in place

~/.claude/skills/judge/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/judge/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about judge

How do I install judge in Claude Code?

Run npx skills add NeoLabHQ/context-engineering-kit --skill judge -a claude-code, run npx skills add NeoLabHQ/context-engineering-kit --skill judge -a claude-code again, then check ~/.claude/skills/judge/SKILL.md.

What does judge do?

The skill extracts work from the current conversation, asks a meta-judge to define review criteria, and sends those criteria to a judge sub-agent working in a separate context. It returns a report without applying changes automatically.

Why does judge use two review stages?

The first stage creates criteria for the specific artifact and review focus. The second stage evaluates the work against those criteria and supports scores with evidence from files and lines.

Similar skills

All skills for Agent setup: memory, context, orchestration →
write-judge-prompt ai-evals-course Walks you through one binary Pass/Fail LLM-as-Judge evaluator for a subjective failure mode: tone, faithfulness, relevance. Useful when code checks fall short. CX Codex +1 1,100installs math-proof-judge official Anthropic A subagent that runs a hard math proof step by step: plans each round's questions, reads the answers, and assembles proof.md. ★ 44,635repo stars affiliate-payout-audit Maya-Beth Finotti Checks one affiliate commission payout run before the money moves: recomputes rates and tiers, catches double-counted conversions, verifies validation windows and tax forms. CX Codex +1 13,500installs partnerships-kickoff Maya-Beth Finotti A router skill for partnerships work: it picks exactly one skill from the 30-skill collection and keeps a shared context file so the next session starts warm. CX Codex +1 13,200installs pricewin-flight-search PriceWin Searches live flight fares across Agoda, Trip.com, and Traveloka for one-way or round trips. It helps compare airline prices and gives direct booking links. CX Codex +1 10,600installs pricewin-booking-assistant PriceWin Recommends hotel rooms, creates OpenTravel bookings with payment links, and provides OTA comparison links. Handles reservation checks and cancellations. CX Codex +1 10,500installs

From the same set64

launch-sub-agent Dispatches a focused sub-agent with isolated context, automated model routing, and mandatory self-critique verification. CX Codex +1 1,200installs do-and-judge This skill sends a task to an implementation sub-agent, checks the result with an independent judge, and retries with feedback when needed. CX Codex +1 1,200installs review-pr Reviews an open GitHub pull request and leaves inline comments on the diff. Handy when the work sits in a PR, not in your local tree. CX Codex +1 1,200installs analyse Selects a Kaizen technique for your target, running code exploration via Gemba Walk, workflow reviews with VSM, or waste audits using Muda. CX Codex +1 1,200installs add-task Turns your request into a task file under .specs/tasks/draft/, with a verb-first title, a type and your original wording. The first step of a spec-driven workflow. CX Codex +1 1,200installs do-in-parallel Runs isolated tasks across separate files concurrently via sub-agents and validates work with automated judges. Fits batch refactoring and doc generation. CX Codex +1 1,200installs analyse-problem This skill turns a problem into a one-page A3 document: background, current condition, goal, root cause and an action plan. CX Codex +1 1,200installs do-in-steps Splits a complex task into dependent steps, runs them in order, and has a separate judge check each step before the next one starts. CX Codex +1 1,200installs

Project sponsors

Companies that keep this analytics open
Gold sponsor
Nutra.Club

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor