# judge-with-debate

The skill runs three independent judges who argue over a solution's score until they agree. Useful when a single pass feels too thin for a real verdict.

- skill · `type=skill` · Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, OpenCode · 1300 installs · ★ 1787
- Author: NeoLabHQ
- License: GPL-3.0 · policy=link-only · description only; the file stays with the author
- Source: https://github.com/NeoLabHQ/context-engineering-kit/tree/HEAD/skills/judge-with-debate/SKILL.md
- Updated: 2026-08-26
- HTML: https://affpapa.org/en/nevibecode/skills/judge-with-debate

## What it does

The skill implements the Multi-Agent Debate pattern. A meta-judge runs once and writes an evaluation specification: rubrics, checklists and a scoring scale. Then three judges analyse the solution in parallel, each on their own, and save reports to `.specs/reports/{solution-name}-{YYYY-MM-DD}.[1|2|3].md`.

Debate rounds follow, up to three of them. Each judge reads the other reports, looks for score gaps wider than one point on any criterion, defends their rating with evidence from the solution and the specification, and revises when the counter-arguments hold. Judges talk only through the filesystem — the orchestrator stays out of it, otherwise its context overflows.

The reports directory `.specs/reports` has to exist before the run; the skill creates it with `mkdir -p .specs/reports`. Both the meta-judge and the judges run on Opus, so expect a heavier bill than a plain review. On a small file the debate mostly adds latency.

## Install

### Claude Code (native platform)

1. **Skills CLI (Vercel)**
```bash
npx skills add NeoLabHQ/context-engineering-kit --skill judge-with-debate -a claude-code
```

2. **Manually: put the files in place**
- File: `~/.claude/skills/judge-with-debate/SKILL.md` — Copy the whole skill folder; for one project use .claude/skills/judge-with-debate/

### Codex

1. **Skills CLI (Vercel)**
```bash
npx skills add NeoLabHQ/context-engineering-kit --skill judge-with-debate -a codex
```

2. **Manually: put the files in place**
- File: `.agents/skills/judge-with-debate/SKILL.md` — For all projects use ~/.agents/skills/

### Cursor

1. **Skills CLI (Vercel)**
```bash
npx skills add NeoLabHQ/context-engineering-kit --skill judge-with-debate -a cursor
```

2. **Manually: put the files in place**
- File: `.cursor/skills/judge-with-debate/SKILL.md` — Cursor also reads .agents/skills/ and .claude/skills/

### GitHub Copilot

1. **GitHub CLI**
```bash
gh skill install NeoLabHQ/context-engineering-kit judge-with-debate
```

2. **Manually: put the files in place**
- File: `.github/skills/judge-with-debate/SKILL.md`

### Gemini CLI

1. **Skills CLI (Vercel)**
```bash
npx skills add NeoLabHQ/context-engineering-kit --skill judge-with-debate -a gemini-cli
```

2. **Platform CLI**
```bash
gemini skills install https://github.com/NeoLabHQ/context-engineering-kit --path skills/judge-with-debate
```

### Windsurf (Devin Desktop)

**Manually: put the files in place**
- File: `.devin/skills/judge-with-debate/SKILL.md` — The legacy .windsurf/skills/ path still works

### OpenCode

**Skills CLI (Vercel)**
```bash
npx skills add NeoLabHQ/context-engineering-kit --skill judge-with-debate -a opencode
```

### Claude.ai

**Claude.ai**
- Zip the folder and upload it in Settings → Capabilities → Skills

Where a file is given instead of a command, copy the files from the source to that path: https://github.com/NeoLabHQ/context-engineering-kit/tree/HEAD/skills/judge-with-debate/SKILL.md

## When to use it
- score a solution against a rubric instead of a gut call
- cut the bias that comes with one reviewer
- surface the disputed criteria and both sides of the argument
- get a report with per-criterion scores and quotes from the code
- compare several solutions against one specification

## Example prompts
- evaluate `src/api/users.ts` with judge-with-debate
- run the judge debate on this module and show where they disagree
- do a multi-agent debate on this file and save the reports to `.specs/reports`

## Similar
- [pre-publish-review](https://affpapa.org/en/nevibecode/skills/pre-publish-review.md) — skill · 753 installs
- [opencode-ensemble](https://affpapa.org/en/nevibecode/skills/opencode-ensemble.md) — skill · 681 installs
- [deep-audit](https://affpapa.org/en/nevibecode/skills/deep-audit.md) — skill · ★ 1668
- [evaluation-rubrics](https://affpapa.org/en/nevibecode/skills/evaluation-rubrics.md) — skill · 784 installs
- [agent-teams](https://affpapa.org/en/nevibecode/plugins/agent-teams.md) — plugin · ★ 41729
- [gem-team](https://affpapa.org/en/nevibecode/plugins/gem-team.md) — plugin · ★ 40418

From the same set: [neolabhq/context-engineering-kit](https://affpapa.org/en/nevibecode/by/neolabhq/context-engineering-kit.md) — 64 cards

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md