ClawHavoc: 341 malicious skills compromised 9,000+ installations

Security Scanner for
Agent Skills

Scan OpenClaw Skills for vulnerabilities, hardcoded secrets, supply chain risks, and more. 20 rules. A-F grading. One command.

Scan a SkillGitHub
terminal
$ clawsafety scan ./my-skill/

  ClawSafety v0.1.0

  Scanning: ./my-skill/
  Findings: 3

  CRITICAL  CS-SEC-002  Hardcoded API Key detected
            scripts/config.py:12
            > api_key = "sk-proj-abc123..."

  HIGH      CS-INJ-001  Shell command injection
            scripts/run.sh:45
            > eval $USER_INPUT

  MEDIUM    CS-DEP-002  Unpinned dependency
            skill.yaml:8
            > requests>=2.0

  Score: 52/100 (D)
  Critical: 1 | High: 1 | Medium: 1

How it works

1

Paste a URL

Enter any GitHub skill repository URL. No login required.

2

Automatic scan

20 rules check for injection, secrets, dependencies, permissions, and config issues.

3

Get your grade

A-F security score with detailed findings and fix suggestions.

20 Security Rules

5 categories covering the full Agent Skill attack surface

INJInjection
4 rules

Shell injection, SQL injection, dangerous functions, reverse shells

SECSecrets
4 rules

Hardcoded passwords, API keys, private keys, URL credentials

DEPDependencies
4 rules

Unsafe installs, unpinned versions, known CVEs, untrusted downloads

PRMPermissions
4 rules

Excessive permissions, sensitive paths, env abuse, insecure chmod

CFGConfig
4 rules

Missing SKILL.md, no version, no permissions, prompt injection

Security Grading

A
90-100
Excellent
B
75-89
Good
C
60-74
Fair
D
40-59
Poor
F
0-39
Dangerous

Secure your skills today

Free for public repositories. No login required.

Scan Now
cargo install clawsafety