CatREFuse

CatREFuse / cls-certify

Public

可能是最好用的 Skill 安全检查 Skill，CocoLoop 出品。

47
3
100% credibility
Found Mar 29, 2026 at 47 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Shell
AI Summary

CLS-Certify is a security certification system that analyzes AI agent skills across six dimensions to assign safety grades from S+ to D.

How It Works

1
📰 Discover a safety checker

You hear about a helpful tool that checks if new AI assistants are safe to use before trying them.

2
📥 Add it to your AI

You simply place this safety checker into your AI's collection of helpers so it's ready whenever you need it.

3
🔍 Spot a new helper

You find an exciting new AI helper online and want to make sure it's trustworthy before adding it.

4
🧐 Ask your AI to scan it

You tell your AI in a casual chat, 'Check if this helper is safe,' and it runs a thorough safety review.

5
📊 Review the safety grade

Your AI shares a clear report with a grade like S or A, highlighting any concerns and tips to stay safe.

✅ Use helpers worry-free

Now you confidently add safe helpers, protecting your setup while enjoying powerful AI tools.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 47 to 47 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is cls-certify?

CLS-Certify is a Shell-based security scanner for AI agent skills—like those for Claude, Copilot, or OpenAI Agents—pulled from GitHub repos or local dirs. It runs six-dimensional checks (static code, dynamic behavior, dependencies, network traffic, privacy, and repo reputation) to detect secrets, prompt poisoning, code execution risks, and more, then grades them S+ to D with Markdown/HTML/PDF reports. Tell your agent "check this GitHub skill anthropic" or "scan my-skill for safety," and it handles batch mode, custom outputs, and configs via natural language or SKILL.md frontmatter.

Why is it gaining traction?

Unlike basic linters, it tailors scans for agent-specific threats like MCP tool abuse or zero-width prompt injections, with agent-verified false positives and radar charts in reports. Devs love the zero-setup install via cocoloop, seamless chat integration for skills from alexa skill github or openclaw skill github, and forced downgrades for hard reds like dynamic code downloads. At 47 stars, it's niche but punches above with MIT license and polished CLI outputs.

Who should use this?

AI skill authors testing github skill tree or skill github kb before publishing. Devs evaluating untrusted github and-skill or github skill antigravity from random repos. Teams auditing installed skills in Claude or Copilot workflows to block D-grade risks.

Verdict

Grab it if you're deep in agent skills—solid for quick certifies despite 1.0% credibility and low stars signaling early maturity. Docs shine, but watch for Shell quirks on non-Unix; pair with manual review for production. Worth the clone for paranoid devs. (198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.