A security-focused CLI tool designed to audit locally installed AI agent skills for exfiltration, execution, and policy risks.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install skill-audit
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install skill-audit using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Skills Audit (SkillLens)?
The Skills Audit skill, powered by the SkillLens CLI, provides a robust framework for identifying security vulnerabilities and policy violations within your skill directories. It scans metadata and bundled resources to produce detailed risk reports, helping developers maintain a secure environment for their automated workflows. By analyzing factors like prompt injection, unauthorized data exfiltration, and arbitrary command execution, this tool ensures that Openclaw Skills remain safe and reliable for production use. It is an essential utility for developers who need to vet third-party or internal skills before they are integrated into sensitive agent environments.
Skills Audit (SkillLens) Use Cases
- Auditing a local repository of Openclaw Skills for potential security vulnerabilities before deployment.
- Identifying skills that request overbroad filesystem or network permissions that could lead to data leaks.
- Verifying the safety of bundled scripts and external asset references to prevent arbitrary code execution.
- Generating risk-focused audit reports for compliance and security reviews in enterprise environments.
- Detecting potential prompt injection or persistence mechanisms that could compromise agent integrity.
How Skills Audit (SkillLens) Works
- The user defines the scan scope by specifying a target directory path or using pre-configured scan roots.
- SkillLens inventories all available skills within the path by parsing the SKILL.md file and associated directory structure.
- The tool executes an automated scan using configured auditors, such as Claude or Codex, to flag immediate security concerns.
- Skills are prioritized based on automated verdicts including safe, suspicious, or unsafe statuses.
- A manual review process is conducted to evaluate specific threats like exfiltration, persistence, and privilege bypass.
- A comprehensive audit report is generated, providing numerical risk scores and actionable evidence-based recommendations.
Skills Audit (SkillLens) Setup
To begin auditing your Openclaw Skills, you can run SkillLens directly via npx or install it as a global CLI tool:
# One-off run
npx skilllens scan
# Global installation
pnpm add -g skilllens
Configure your scan roots and check for available auditors using the config command:
skilllens config
Skills Audit (SkillLens) Data Schema & Taxonomy
The skill organizes its findings by analyzing the internal structure of each skill directory. The following data points are evaluated:
| Component | Description |
|---|---|
| SKILL.md | Analyzes primary instructions for prompt injection and overbroad triggers. |
| scripts/ | Inspects bundled shell commands or binaries for execution risks. |
| references/ | Validates external documentation or data files for suspicious links. |
| assets/ | Scans media or binary assets for hidden payloads. |
All findings are consolidated into a report featuring a verdict, a risk score (0-100), and specific evidence quoted from the source files.
Skills Audit (SkillLens) Advanced Features
- Support for multiple auditor CLIs including Claude and Codex for cross-model security verification.
- Verbose output mode to inspect the raw reasoning of the auditor during the scan process.
- Cache-bypass functionality using the --force flag to ensure fresh analysis of updated Openclaw Skills.
- Specific path targeting to isolate scans to newly downloaded or experimental skill directories.
- Risk evaluation focus on realistic abuse scenarios like SSH key exfiltration and shell profile persistence.
SKILL.md
Loading
Related Openclaw Skills
Chromecast Control
A comprehensive CLI-based skill for discovering, casting media, and managing playback on Chromecast and Google Cast-enabled devices.
Tempo Workspace
A professional integration that connects Openclaw Skills to Tempo for real-time workspace awareness and automated insight extraction.
OpenClaw Config Backuper
An automated utility to backup OpenClaw configurations, local skills, and workspace data for disaster recovery and migration.
Moltbook User
A professional skill for AI agents to interact with the Moltbook social network through bilingual content management and secure API operations.
Council
A sophisticated multi-model deliberation tool that gathers independent analysis from diverse AI providers to generate a synthesized consensus on complex problems.
Icom IC-7610 Transceiver Control
A comprehensive automation interface for the Icom IC-7610 transceiver, enabling remote control, CW keying, and real-time telemetry monitoring.
Featured*
