Skill Sentinel - Skill Security Scanner

Live

A watchtower for Skill security before install time

OpenClawAudit

Skill Sentinel - Skill Security Scanner

Published on ClawHub as skill-sentinel.

Protect your OpenClaw environment from malicious Skills.

The Problem

ClawHub is open for anyone to upload Skills. This creates a supply chain attack surface:

  • Malicious scripts disguised as helpful tools
  • Social engineering tricks ("please copy and run this command")
  • Hidden network callbacks
  • Obfuscated code

The Solution

Skill Sentinel scans Skills before installation:

  1. Static Analysis - Examine SKILL.md and scripts without execution
  2. Pattern Detection - Match against known attack signatures
  3. Risk Scoring - Clear Safe/Caution/Avoid ratings
  4. Evidence - Specific line numbers and explanations

Risk Categories

LevelDescriptionAction
🟒 SafeNo suspicious patterns foundInstall freely
🟑 CautionSome risky patterns detectedReview before install
πŸ”΄ AvoidHigh-risk patterns detectedDo not install

Integration

# CLI usage
npx clawshield scan ./my-skill

# GitHub Action
- uses: bubustudio/clawshield-action@v1

Roadmap

  • Core scanner engine
  • CLI tool
  • GitHub Action
  • IDE extension
  • Real-time ClawHub integration
Trust LayerΒ·Community Profile
Trust Layer
Data handling is explicit
Open and auditable workflow
Runtime boundaries are transparent
Risk Note

Third-party skills may include hidden network calls or unsafe install scripts. Scan before use.