benlee-skillguard
bởi benlee2144
Coding Agents & IDEsTóm tắt
Security scanner that audits OpenClaw skills for malicious code, prompt injection, supply chain attacks, data.
Mô tả
Kỹ năng này làm gì
Security scanner that audits OpenClaw skills for malicious code, prompt injection, supply chain attacks, hardcoded secrets, and data exfiltration. Uses a point-based scoring system with three risk levels and maintains SHA-256 file baselines to detect tampering between scans. Runs entirely on Python 3 stdlib with no external dependencies.
Generic code scanners don't understand the OpenClaw skill format or prompt injection patterns, while SkillGuard includes an 80+ domain allowlist and context-aware scoring that cuts false positives for skills making legitimate API calls.
Khi nào nên dùng
- Auditing all installed skills before running them on a new machine
- Checking a newly downloaded skill from ClawHub before trusting it
- Running scheduled cron alerts when a skill's files change unexpectedly
- Detecting hardcoded AWS or GitHub tokens inside third-party skills
- Identifying typosquatted package names in skill dependencies
Quy trình ví dụ
User asks: scan all my installed skills and write a report
- 1
Run skillguard.py scan against ~/clawd/skills/
- 2
Compute SHA-256 baselines for each skill's files on first run
- 3
Score each skill based on detected patterns, domain analysis, and sensitive file access
- 4
Flag any skills that exceed thresholds or trigger automatic MALICIOUS conditions
- 5
Write a markdown report with risk levels and per-finding recommendations
Markdown report listing each skill as CLEAN, SUSPICIOUS, or MALICIOUS with one-line recommendations for every finding
Cài đặt
clawhub install benlee2144/benlee-skillguardChạy lệnh CLI, hoặc dán prompt vào Claude Code, Cursor, Codex hay coding agent bạn đang dùng.