Skill Security Auditor

Skill Security Auditor

Metadata

• Name: skill-security-auditor
• Version: 1.1.0
• Author: Galatea / OpenClaw Team
• Category: Security
• Description: OpenClaw skill security auditor with Cisco AI Skill Scanner integration

Requirements

• Python 3.8+
• Cisco AI Skill Scanner (pip install cisco-ai-skill-scanner[all])
• Dependencies in requirements.txt

Installation

pip install cisco-ai-skill-scanner[all]

Usage

As CLI Tool

# Scan single skill
./tools/claw-audit.py scan <skill-path>

# Scan all skills
./tools/claw-audit.py scan-all

# Check status
./tools/claw-audit.py status

As Python Module

from lib.scanner_orchestrator import ScannerOrchestrator

orchestrator = ScannerOrchestrator()
result = orchestrator.scan_skill("/path/to/skill")

for finding in result.findings:
    print(f"{finding.severity}: {finding.message}")

Commands

scan

Scan a single skill for security issues.

Arguments:

• skill_path: Path to skill directory

Options:

• -s, --scanners: Scanners to use (cisco, clawsec)
• --severity: Minimum severity (critical, high, medium, low, info)
• --no-cve: Skip CVE database check
• --auto-remediate: Auto-quarantine critical issues
• -o, --output: Output file path
• -f, --format: Output format (sarif, json, markdown)
• --stdout: Print to stdout
• -d, --detailed: Show detailed findings

scan-all

Scan all skills in a directory.

Arguments:

• skills_dir: Skills directory (optional, uses OPENCLAW_SKILLS_DIR)

Options:

• -w, --workers: Parallel workers (default: 4)
• --severity: Minimum severity
• --auto-remediate: Auto-quarantine

status

Check scanner status and availability.

Output Formats

SARIF 2.1.0

GitHub Code Scanning compatible format.

Markdown

Human-readable report with severity icons and recommendations.

JSON

Machine-readable format for programmatic consumption.

Threat Categories

AITech classification system:

1. Prompt Injection
2. Data Exfiltration
3. Credential Harvesting
4. Command Injection
5. Dependency Confusion
6. Malicious Code Execution
7. Network Egress
8. Privilege Escalation
9. Obfuscation
10. Backdoor
11. Supply Chain Attack

Configuration

Environment Variables

• OPENCLAW_SKILLS_DIR: Default skills directory
• CISCO_SCANNER_THRESHOLD: Severity threshold
• CLAWSEC_FEED_URL: Threat feed URL
• AUDIT_AUTO_REMEDIATE: Enable auto-remediation

Config Files

• config/default.yaml: Default settings
• config/cve_cache.json: CVE database cache

Exit Codes

Development

Running Tests

pytest tests/ -v --cov=lib

Adding New Scanners

1. Implement scanner class in lib/
2. Add to ScannerOrchestrator
3. Update SPEC.yaml
4. Add tests

Changelog

1.1.0 (2026-02-11)

• Migrated from AgentVerus to Cisco AI Skill Scanner
• Unified Python stack (removed Node.js dependency)
• AITech threat classification
• SARIF 2.1.0 output format

1.0.0

• Initial release with AgentVerus
• Basic static analysis
• ASST taxonomy