Skill Vetter - Setup Guide¶
Source: useai-pro/openclaw-skills-security (Community)
Skill: skill-vetter · Installs: 20.6K+ · Category: Security
Platform: Linux, macOS, Windows
Skill Vetter is a security-first pre-install audit tool for OpenClaw skills. Before installing any skill from ClawHub, GitHub, or other sources, it runs a structured red-flag checklist covering permissions requested, suspicious instruction patterns, file access scope, and network/shell access requirements. With a 97% trust score and last audited February 2026.
Installation¶
npx skills add useai-pro/openclaw-skills-security@skill-vetter
Prerequisites¶
| Requirement | Details |
|---|---|
| OpenClaw or Hermes Agent | Skill execution environment |
| Skill to vet | A SKILL.md from ClawHub, GitHub, or shared source |
When to Use¶
- Before installing a new skill from ClawHub
- When reviewing a SKILL.md from GitHub or other sources
- When someone shares a skill file and you need to assess safety
- Before integrating community skills into production workflows
Vetting Checklist¶
The skill audits against these red-flag categories:
| Category | What It Checks |
|---|---|
| File Access | Does the skill read/write files? Which directories? |
| Network Access | Does it make outbound requests? To which domains? |
| Shell Access | Does it execute shell commands? With what privileges? |
| Permission Scope | What allowed-tools are declared? Are they over-broad? |
| Suspicious Patterns | Obfuscated code, encoded payloads, eval/exec usage |
| Data Exfiltration | Does it send data off-machine? To where? |
| Dependency Risk | External package installs, version pinning, supply chain |
Output¶
Produces a conservative manual review with:
- Trust Score - overall safety assessment
- Red Flags - specific concerns with line references
- Permission Summary - what the skill can access
- Recommendation - Install, Review Further, or Block
Trust Score: 97/100¶
The skill itself has been audited: - Author: useclawpro - Category: Security - File Read: Yes (required for auditing) - File Write: No - Network: No - Shell: No - Last Audited: February 2026
Verification¶
After vetting a skill: - Review the red-flag report before installing - Cross-reference permissions against the skill's stated purpose - For production, require second reviewer for any skill with shell or network access
Related Skills¶
Integration with CorpusIQ¶
CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.
This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.