Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
What This Skill Does
Scans OpenClaw skill directories for supply chain attacks and malicious code using 22 detection rules. Context-aware analysis distinguishes documentation from executable code to reduce false positives. Outputs colored terminal results or JSON reports.
Zero external dependencies means it runs on any macOS or Linux system without setup, unlike scanners that require language runtimes or package installs.
When to use it
- Auditing a new third-party skill before installing it
- Running automated security checks in a CI pipeline for skill repos
- Identifying prompt injection attempts hidden in SKILL.md files
- Detecting typosquatted npm/pip packages bundled with skills
- Finding reverse shells or credential exfiltration patterns in skill scripts
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: scan the skills directory and give me a JSON report
- 1Run audit.sh with --json flag pointing to the target skills directory
- 2Scanner applies all 22 detection rules across each skill's files
- 3Context-aware engine filters out rule mentions inside documentation
- 4Assigns severity levels (critical/warning) to each finding
- 5Writes JSON report with matched lines, rule IDs, and file paths
JSON report listing all critical and warning findings with file locations, matched content, and rule identifiers, plus exit code 2 if critical issues were found