# Judge

> Mandatory pre-build challenge gate. Use before approving or building any feature, connector, automation, agent, hook, MCP server, UI change, database change, or architecture plan. Performs first-source evidence review, devil's advocate critique, existing capability review, UX review, security/privacy review, test/stress review, and return-on-effort scoring.

- Skill: `cleanexpo/judge-2` (Agent Skill)
- Install (CLI): `npx skillmds@latest add cleanexpo/judge-2`
- Raw SKILL.md: https://api.skillmd.com/api/skills/cleanexpo/judge-2/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: cleanexpo (https://skillmd.com/u/cleanexpo)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/cleanexpo/judge-2

---


# judge — First Evidence Challenge Gate

Run this skill before any build, implementation, connector, hook, agent, MCP server, database change, UI change, or architecture plan is approved.

Judge this proposal:

```text
$ARGUMENTS
```

## Hard rule

Do not build. Do not edit. Do not commit. Do not push. Do not deploy.

This is a read-only pre-build review.

## Required review passes

- First-source evidence review
- Existing capability review
- Devil's advocate challenge
- Architecture and bloat review
- Security and privacy review
- UI/UX friction review
- Test, loop, and stress review
- Return-on-effort scoring

## Evidence ranking

Use first-source evidence wherever possible:

1. Official docs
2. Official SDK/API references
3. Official changelogs
4. Repo source code
5. Tests, CI, logs, traces, schemas, migrations
6. Standards/specs
7. Known expert material
8. Blogs/videos/social only as discovery

LLM memory is not evidence.

Unsupported claims must be marked `UNSUPPORTED`.

## Score

Score out of 100:

| Category | Weight |
|---|---:|
| First-source evidence | 25 |
| Clear user/business problem | 20 |
| Reuse of existing capability | 15 |
| Security/privacy safety | 15 |
| UX clarity | 10 |
| Testability | 10 |
| Cost/control simplicity | 5 |

Decision rules (HARD LINE — 100/100 is the only build bar):

- **APPROVE BUILD requires a real 100/100** — every mandatory Convergence criterion satisfied.
  There is no 85 pass; 85–99 is a list of gaps to close, not an approval.
- 0–99 = **NOT APPROVED.** Iterate per Convergence until a real 100 is earned, or halt and
  report the honest ceiling with the exact blocker. Never approve below 100, never inflate.
- REJECT (0–69) / REDUCE SCOPE / APPROVE EXPERIMENT (70–99) are iteration states only; none
  authorise a production build. Only a real 100 does.

## Convergence — do not stop until a REAL 100/100

`/judge` does not end at the first score. It iterates: score → list every gap with its
first-source anchor → drive the real fix (gather the missing evidence, reduce or reshape
scope, clean cache and bloat, correct any false claim) → re-score. Repeat until the proposal
genuinely earns 100/100.

A 100 is valid ONLY when ALL of these hold — never by inflation:
- **Real data:** every evidence row is SUPPORTED by first-source; zero UNSUPPORTED / PARTIAL /
  NOT CHECKED remain. Checked, not asserted.
- **Cache and bloat cleaned:** no dead code, duplication, sediment, unused abstraction, or
  stale copy survives the deletion test.
- **True and correct:** every claim is verified against the source at real scale — no
  plausible-but-unproven statement counts.
- **No open blocker:** all review passes clear.

**Honesty rail (non-negotiable):** if the scope cannot honestly reach 100, do NOT fake the
number — reshape it until a real 100 is reachable, or halt and report the honest ceiling with
the exact reason and what would lift it. A fabricated 100 is a gate failure, not a pass.

## Output format

### Judge Report

1. Proposal being judged
2. Decision
3. Score
4. First-source evidence table
5. What already exists
6. Devil's advocate objections
7. Architecture and bloat risks
8. Security, privacy, and permission risks
9. UI/UX missing elements
10. Loop testing and stress testing
11. Smallest safe version
12. Final recommendation