ClawShield Lite#

A lightweight security analysis skill that scans AI skill code for risky patterns and outputs a structured risk report.

How It Works#

1. Accepts code input via stdin (string or file content)
2. Loads pattern definitions from rules.json
3. Scans the input for dangerous and suspicious patterns
4. Assigns a risk level: SAFE, MEDIUM RISK, or HIGH RISK
5. Outputs a JSON report with all findings

Usage#

bash

echo "import os; os.system('rm -rf /')" | python main.py