luiscamaral

skill-reviewer

1
0
# Install this skill:
npx skills add luiscamaral/k8s-cell-platform-skills --skill "skill-reviewer"

Install specific skill from multi-skill repository

# Description

Reviews and validates Claude Code skill definitions for best practices, non-overlapping domains, and correct configuration. Use when creating new skills, reviewing existing skills, checking for conflicts, or validating skill quality.

# SKILL.md


name: skill-reviewer
description: Reviews and validates Claude Code skill definitions for best practices, non-overlapping domains, and correct configuration. Use when creating new skills, reviewing existing skills, checking for conflicts, or validating skill quality.
allowed-tools: Read, Glob, Grep


Skill Reviewer

Validates Claude Code skills for the Kubernetes Cell Platform project.

Validation Checklist

1. Frontmatter Validation

  • [ ] name is kebab-case
  • [ ] name is ≤64 characters
  • [ ] description is ≤1024 characters
  • [ ] description includes trigger keywords
  • [ ] description describes WHAT and WHEN to use
  • [ ] allowed-tools matches execution mode

2. Content Validation

  • [ ] SKILL.md is <500 lines
  • [ ] Clear step-by-step instructions
  • [ ] Memory file paths are correct
  • [ ] Reference docs exist if linked
  • [ ] Scripts exist and are executable

3. Non-Overlap Verification

Each skill must have a clear, non-overlapping domain:

Skill Domain Exclusive Keywords
platform-overview Status/health overview, status, deployed
l0-infrastructure Terraform/Talos terraform, talos, bootstrap, cilium
l1-cluster-services L1 components metallb, external-dns, karpenter
l2-gitops-platform L2 components argocd, kyverno, linkerd, gitops
troubleshooting-pods Pod debugging crash, logs, debug, oom, pending
security-audit Security review rbac, audit, security, cve

4. Execution Mode Check

Mode Allowed Tools Pattern
Read-only No write verbs in Bash
Tiered Bash(cmd:read-verbs) for auto, others need approval

Review Process

Step 1: Parse Frontmatter

# Extract frontmatter from SKILL.md
head -20 .claude/skills/<skill-name>/SKILL.md

Step 2: Check for Overlaps

Compare trigger keywords across all skills:

# List all skill descriptions
for skill in .claude/skills/*/SKILL.md; do
  echo "=== $skill ==="
  grep "^description:" "$skill"
done

Step 3: Validate References

# Check if referenced files exist
grep -E "reference/|scripts/" .claude/skills/<skill-name>/SKILL.md | while read ref; do
  # Verify file exists
done

Step 4: Line Count

wc -l .claude/skills/*/SKILL.md

Best Practices

Description Writing

Good:

description: Manages Talos Linux and Terraform infrastructure layer (L0). Use for terraform operations, talos cluster management, node provisioning, or cluster bootstrap.

Bad:

description: Helps with infrastructure  # Too vague

Tool Restrictions

# Read-only skill
allowed-tools: Read, Glob, Grep, Bash(kubectl:get,describe,logs)

# Tiered skill (some ops need approval)
allowed-tools: Read, Glob, Grep, Bash(kubectl:*, terraform:*)

Progressive Disclosure

SKILL.md           # Quick reference (<500 lines)
├── reference/     # Detailed docs (loaded on demand)
└── scripts/       # Utilities (executed, not read)

Reference Documentation

  • reference/skill-checklist.md - Detailed validation steps
  • reference/best-practices.md - Skill authoring guide

Common Issues

Issue Solution
Skill not triggering Description too vague, add specific keywords
Overlap with another skill Narrow the domain, update description
Memory file not found Create file or update path
Script not executing Check chmod +x and path format

# Supported AI Coding Agents

This skill is compatible with the SKILL.md standard and works with all major AI coding agents:

Learn more about the SKILL.md standard and how to use these skills with your preferred AI coding agent.