All skills
simota avatar

/judge

@e307415
by shingo imotasimota/agent-skills85 stars
15

Reviewing code via multi-engine orchestration (Claude + Codex) on three axes — secure, correct, and lean — shipping only findings worth fixing. Use for PR review or pre-commit. Complements Zen.

Use this Skill: https://skilld.dev/gh/simota/agent-skills/judge

This session only. Nothing lands on disk.

referenceskill-review-criteria.md

≈1.5k tokens on demand. Your agent reads this file only when SKILL.md points to it.

Skill Review Criteria Reference

Source: "The Complete Guide to Building Skills for Claude" (Anthropic, 2025)

Official criteria reference that Judge consults when reviewing SKILL.md files.


1. Determining SKILL.md Review Scope

Conditions for Review

Apply SKILL.md review mode when any of the following holds:

  • The review target includes a SKILL.md file
  • The review target includes a reference/*.md file (within a skill)
  • A .md file with YAML frontmatter was changed
  • A change occurred under the skills/ directory

How This Differs from Regular Code Review

Aspect Code Review SKILL.md Review
Primary Tool codex review CLI Official criteria checklist
Focus Correctness, security, logic Structure, description quality, progressive disclosure
Severity Scale CRITICAL-INFO (5 levels) CRITICAL-LOW (4 levels, based on official criteria)
Routing Builder / Sentinel / Zen Sigil / Architect / Gauge

2. SKILL.md Review Checklist

Critical (Blocking)

ID Check Rule
SK-C01 SKILL.md filename is exact (case-sensitive) Only SKILL.md is allowed
SK-C02 YAML frontmatter --- delimiters are correct Both opening and closing are required
SK-C03 A name field is present Required field
SK-C04 A description field is present Required field
SK-C05 No XML tags (< >) in frontmatter Security restriction
SK-C06 name does not contain "claude" / "anthropic" Reserved prefix

High (Must Fix)

ID Check Rule
SK-H01 name is kebab-case Regex: ^[a-z0-9]+(-[a-z0-9]+)*$
SK-H02 name matches the folder name String match
SK-H03 description is 1024 characters or fewer Official limit
SK-H04 description states WHAT Describes what the skill does
SK-H05 description states WHEN Trigger conditions for when to use it
SK-H06 description is not vague Excludes "Helps with projects"-level phrasing
SK-H07 A step-by-step structure exists Heading or numbered list

Medium (Recommended Improvement)

ID Check Rule
SK-M01 An Examples section exists Concrete user-scenario examples
SK-M02 A Troubleshooting section exists Documents error handling
SK-M03 The reference/ directory is utilized Progressive Disclosure
SK-M04 SKILL.md is 5000 words or fewer Context efficiency
SK-M05 Critical instructions appear near the top of the document High-priority instructions placed first

Low (Optional Improvement)

ID Check Rule
SK-L01 compatibility field is 500 characters or fewer Official limit
SK-L02 A negative trigger exists (where needed) Overtriggering prevention
SK-L03 metadata includes author / version Best practice
SK-L04 No README.md exists in the skill folder Official rule

3. Description Quality Assessment

Good Description Pattern

# Pattern: Specific + Actionable + Trigger phrases
description: Analyzes Figma design files and generates developer handoff
documentation. Use when user uploads .fig files, asks for "design specs",
"component documentation", or "design-to-code handoff".

Checkpoints:

  • ✅ Starts with a verb (clear about what it does)
  • ✅ Includes trigger phrases (clear about when to use it)
  • ✅ Mentions file types (clear about the target)

Bad Description Anti-Patterns

Anti-pattern Example Issue
Too vague "Helps with projects" Fails to trigger
Missing triggers "Creates documentation systems" No WHEN
Too technical "Implements entity model with hierarchical relationships" Not written from the user's perspective
Too long (>1024) — Bloats the frontmatter

Verdict Logic

IF any SK-C* fails → BLOCK (CRITICAL findings)
IF SK-H04 OR SK-H05 fails → REQUEST CHANGES
IF 3+ SK-M* fail → REQUEST CHANGES (accumulation)
ELSE → APPROVE (with notes if SK-L* issues exist)

4. Progressive Disclosure Review

Verifying the Three-Level Structure

Level What to Check Finding if Missing
1st (Frontmatter) Are name + description minimal yet sufficient? SK-H04, SK-H05
2nd (Body) Does SKILL.md stay focused on core instructions? SK-M04 if word count > 5000
3rd (References) Are details separated out into reference/? SK-M03

Context Efficiency Assessment

IF SKILL.md > 5000 words AND reference/ is empty:
  → Finding: SK-M03 + SK-M04
  → Recommendation: "Move detailed documentation to reference/"

IF SKILL.md < 500 words AND reference/ has 5+ files:
  → Note: Good progressive disclosure structure

IF all content is inline AND no reference/ directory:
  → Finding: SK-M03
  → Recommendation: "Consider progressive disclosure structure"

5. Review Report Format

SKILL.md Review Finding

ID: [SK-C/H/M/L + NN]
Severity: [CRITICAL | HIGH | MEDIUM | LOW]
Location: [SKILL.md:line or structural]
Standard: Anthropic Agent Skill Specification (2025)
Finding: [specific observation]
Recommendation: [actionable fix]
Remediation: [Sigil | Architect | Gauge]

Section to Add to the Summary

## Skill Quality Assessment
- Frontmatter: [PASS | FAIL] ([count] issues)
- Description Quality: [PASS | FAIL] ([count] issues)
- Instruction Structure: [PASS | FAIL] ([count] issues)
- Progressive Disclosure: [PASS | FAIL] ([count] issues)
- Overall: [COMPLIANT | PARTIAL | NON-COMPLIANT]

6. Routing

Finding Type Route To
Frontmatter structure issue Sigil (regenerate) or Architect (redesign)
Description quality issue Sigil (improve description)
Instruction structure issue Sigil (restructure)
Progressive Disclosure issue Architect (design reference separation)
Ecosystem internal standards violation Gauge (16-item checklist)

Source: SKILL.md on GitHub

No alerts13d5 checks · Risk SAFE
  • Gen Agent Trust Hub13d

    The 'judge' skill is an advanced code review tool designed to analyze software changes using multiple AI engines including Claude, Codex, and Google Gemini. It features sophisticated workflows for PR reviews, security audits, and performance checks. While it processes external data like code diffs and PR descriptions, it includes a multi-layered verification process to ensure findings are accurate. The skill also includes technical instructions for managing CLI tool requirements, such as handling terminal interactions using Python. All external resources and download instructions point to trusted organizations like Google and Anthropic.

  • Socket13d

    No alerts

  • Snyk13d

    Risk: LOW · No issues

  • Runlayer6mo

    1/11 files flagged

  • ZeroLeaks5mo

    Score: 93/100 · 2 sections analyzed

Signed by skilld at e307415. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 2 days ago.

Activeupdated 2 weeks ago

README badge

README badge for simota/agent-skills/judge