🔒 Skill Scanner (v1.1.0)

Enterprise-Grade Security Orchestrator for AI Agent Skills

Skill Scanner is a powerful security tool designed to protect your AI ecosystem by validating Agent Skills against advanced threats, "Shadow AI" patterns, and supply chain attacks.

🚀 Quick Start (No Installation Needed)

You can run Skill Scanner instantly on any repository or file using npx:

# Basic scan
npx skill-scanner ./my-skills

# Scan with full Enterprise features (AI + Malware Intelligence)
export GEMINI_API_KEY="your_api_key"
npx skill-scanner ./my-skills --report --sarif

🔥 Enterprise Features

• 🧠 AI Semantic Analysis: Uses Gemini 2.0 to detect malicious intent, data exfiltration, and Prompt Injection (Jailbreak attempts).
• 🛡️ SARIF Support: Ready for GitHub Security Tab. Export standard reports for enterprise audit trails.
• 📊 Risk Scoring (A-F): Instant assessment of your skill's risk level with weighted category analysis.
• 📦 Dependency Audit: Scans package.json for deprecated, malicious, or typosquatted packages.
• 🔗 VirusTotal Integration: Reality-check code hashes against global malware intelligence.
• 🏷️ OWASP LLM Top 10 Mapping: Every finding is automatically mapped to industry-standard LLM risk categories.

🛠️ Usage

Installation

If you prefer to install it globally:

npm install -g skill-scanner

Advanced Scan Examples

# 1. Standard scan with Risk Assessment
skill-scanner ./skills

# 2. Automated Report generation (JSON + SARIF)
skill-scanner ./skills --report --sarif

# 3. High-Security mode (only report High/Critical)
skill-scanner ./skills --severity high

# 4. Filter specific checks
skill-scanner ./skills --checks semantic-analysis,dependency-audit

🤖 GitHub Actions Integration

Automate your security checks by adding our ready-to-use workflow. Create .github/workflows/skill-scan.yml:

uses: JonusNattapong/Skill-Scanner/.github/workflows/skill-scan.yml@main
with:
  api-key: ${{ secrets.GEMINI_API_KEY }}

This will automatically upload results to your GitHub Security results (SARIF).

🏷️ Risk Coverage

| Category | OWASP Mapping | Description | |----------|---------------|-------------| | Prompt Injection | LLM01 | Detection of jailbreak patterns and instruction overrides. | | Data Exfiltration | LLM02 | DNS tunneling, unauthorized webhooks, and credential theft. | | Supply Chain | LLM03 | Malicious dependencies and typosquatted package names. | | Excessive Agency | LLM06 | Hidden backdoors, reverse shells, and unauthorized persistence. | | Command Injection | LLM05 | Unsafe execution of system commands. |

📄 License

Distributed under the MIT License. See LICENSE for more information.

Developed by JonusNattapong for the Secure AI Future.