skills.rest
Evol-ai
Evol-aiOFFICIAL
Explore all skills by this author →
GitHub

skill-compass

Scores six dimensions of Claude Code/OpenCLaw skills and identifies the weakest for improvement.

229|9|Updated Mar 19, 2026
Search Tags:#usage tracking#skill quality#security gate#directed improvement#local-first evaluation#version snapshots
One-click install
npx skills add https://github.com/Evol-ai/SkillCompass --skill skill-compass
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: skill-compass
Source: https://github.com/Evol-ai/SkillCompass/tree/main
Command: npx skills add https://github.com/Evol-ai/SkillCompass --skill skill-compass

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires js-yaml, typescript, and includes scripts (resource) and assets (resource) components.

What problem does it solve?

SkillCompass evaluates Claude Code / OpenClaw skills across structure, trigger quality, and security, then guides you to fix the weakest dimension while verifying the improvement.

Core Features & Use Cases

  • Six-dimension skill evaluation: scores D1–D6 and produces a quality verdict (with D3 security as an absolute gate).
  • Directed improvement with verification: runs eval-improve in targeted rounds (snapshot before write, then targeted re-eval to confirm improvement).
  • Usage-driven Skill Inbox: tracks skill usage locally and surfaces suggestions for idle, stale, risky, or never-evaluated skills.
  • Version tracking and rollback: snapshots, manifests, and audit logs support recovery if improvements regress quality.

Quick Start

Run /eval-skill on a skill path (for example, /eval-skill ./my-skill/SKILL.md) to get a report showing which dimension is the weakest and what to fix next.

Frequently Asked Questions about skill-compass

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I evaluate Claude Code skill quality and find what to fix?▼

You evaluate Claude Code skill quality by scoring six dimensions from D1 Structure to D6 Uniqueness, which identifies the weakest area for prioritized fixes and deterministic verification.

What is local-first usage tracking for stale or risky skills?▼

Local-first usage tracking monitors skill activity locally to surface a Skill Inbox recommending fixes or removals for idle, stale, risky, or never-evaluated skills.

How do I make directed improvements to a skill and verify the changes?▼

Run targeted eval-improve rounds that take a version snapshot before writing changes, then perform a targeted re-evaluation to verify the quality improvement without regressions.

Can I run batch audits on multiple skills at once?▼

Yes, batch audits support applying six-dimension skill evaluation across multiple installed skills simultaneously to generate comprehensive quality reports and directed fix recommendations.

Does skill evaluation enforce security checks before scoring functionality?▼

Yes, D3 Security acts as an absolute gate during skill evaluation, meaning skills must pass security checks before functional, comparative, and uniqueness dimensions are fully scored.

How do version snapshots and manifests support rollback for skill improvements?▼

Version snapshots, manifests, and audit logs capture the state before writing improvements, supporting immediate recovery and rollback if directed fixes regress overall skill quality.

Related Skills

Explore More Software Engineering Skills→
mkgs-databricks-demosmkgs-databricks-demos
1

databricks-mlflow-evaluation

Locate root SKILL.md files and extract YAML frontmatter metadata into YAML documents.

Official
Advanced
antoniolgantoniolg
97

email

Aggregate Gmail and iCloud inbox metadata for AI-driven triage.

Community
Intermediate
z1439527767z1439527767

skill-judge

Evaluate installed AI skills across trigger accuracy, resolution, efficiency, and consistency.

Community
Intermediate
OneWave-AIOneWave-AI
245

code-review-pro

Review code for security vulnerabilities, performance bottlenecks, and maintainability issues.

Official
Basic