HN Showcase
Back to browse
GitHub Repository

Evaluate agent skill quality. Find the weakest link. Fix it. Prove it worked.

224 starsJavaScript

SkillCompass – Diagnose and Improve AI Agent Skills Across 6 Dimensions

by yo103jg·Apr 3, 2026·2 points·0 comments
Visit ProjectView on HN

AI Analysis

●●SolidBig BrainNiche Gem

Six-dimension scoring framework beats guesswork for improving Claude Code skills.

Strengths
  • •Structured 6-dimension scoring (structure, trigger, security, functional, comparative, uniqueness)
  • •Iterative workflow finds weakest dimension, fixes it, verifies, then moves to next
  • •Local Node.js validators run security checks without sending code externally
Weaknesses
  • •Narrow audience — only developers building Claude Code or OpenClaw skills
  • •Scoring methodology depends on Claude Opus 4.6 consistency — unproven at scale
Category
AI/ML
Target Audience

Developers building and maintaining AI agent skills

Similar To

Claude Code · OpenClaw · Cursor

Post Description

SkillCompass is an evaluation-driven skill evolution engine for Claude Code and OpenClaw. It scores your Claude Code and OpenClaw skills across 6 dimensions (structure, trigger, security, functional, comparative, uniqueness), pinpoints the weakest one, and fixes it, then moves to the next weakest. It also detects when model improvements make a skill unnecessary. Runs locally. Requires Node.js v18+ for local validators.

Similar Projects