ClawKit Logo
ClawKitReliability Toolkit
Back to Registry
Official Verified

guardskills

Security gate for skill installs. Scans skill content for malicious patterns (credential exfiltration, remote code execution, destructive ops, privilege escalation) before installation. Always run guardskills before installing any skill from ClawHub, GitHub, or local sources.

skill-install — Terminal

Install via CLI (Recommended)

clawhub install openclaw/skills/skills/felixondesk/guardskills
Or

GuardSkills — Security Scanner for Skills

Static analysis security gate. Scans skill files for malicious patterns and produces a risk decision before allowing installation.

Rule — ALWAYS scan before installing

Before installing any skill (ClawHub, GitHub, local, or skills.sh), you MUST run a guardskills scan first. Never run clawhub install, skills add, or copy skill files without scanning.

Scan by source

1. ClawHub skills

When the user asks to install a ClawHub skill, scan it first:

guardskills scan-clawhub <owner/skill-slug>

Also accepts full URLs:

guardskills scan-clawhub https://clawhub.ai/owner/skill-slug

2. Local skills

When loading or reviewing a skill folder on disk:

guardskills scan-local <path-to-skill-folder>

If the path contains multiple skills:

guardskills scan-local <path-to-skills-root> --skill <skill-folder-name>

3. GitHub skills

When installing from a GitHub repository:

guardskills add owner/repo --skill <skill-name> --dry-run

Also accepts full URLs:

guardskills add https://github.com/owner/repo --skill <skill-name> --dry-run

4. skills.sh skills

Use the same GitHub flow:

guardskills add owner/repo --skill <skill-name> --dry-run

If scan passes:

guardskills add owner/repo --skill <skill-name>

How to interpret results

DecisionAction
SAFEProceed with installation.
WARNINGAsk the user for confirmation before proceeding.
UNSAFEBlock. Tell the user the skill is unsafe and do not install.
CRITICALBlock. Do not install under any circumstances.
UNVERIFIABLEBlock. Inform the user the skill could not be verified.

Useful flags

  • --json — machine-readable JSON output
  • --dry-run — scan only, do not install
  • --ci — deterministic gate mode (no prompts, no install)
  • --strict — use stricter risk thresholds
  • --yes — auto-accept WARNING level (only if user explicitly requests)
  • --force — override UNSAFE block (only if user explicitly requests)
  • --allow-unverifiable — override UNVERIFIABLE block (only if user explicitly requests)

Exit codes

  • 0 — allowed / safe
  • 10 — warning not confirmed
  • 20 — blocked (UNSAFE, CRITICAL, or UNVERIFIABLE)
  • 30 — runtime / internal error

Recommended workflow

  1. User asks to install a skill.
  2. Determine the source (ClawHub, GitHub, local, skills.sh).
  3. Run the appropriate guardskills scan command.
  4. Read the decision from the output.
  5. If SAFE, proceed with clawhub install or equivalent.

Metadata

Stars2387
Views0
Updated2026-03-09
View Author Profile
AI Skill Finder

Not sure this is the right skill?

Describe what you want to build — we'll match you to the best skill from 16,000+ options.

Find the right skill
Add to Configuration

Paste this into your clawhub.json to enable this plugin.

{
  "plugins": {
    "official-felixondesk-guardskills": {
      "enabled": true,
      "auto_update": true
    }
  }
}
Safety NoteClawKit audits metadata but not runtime behavior. Use with caution.