skillaudit.sh

Security Checks

What we scan for in LLM skill files

skillaudit-data-exfiltrationcritical

Data Exfiltration Pattern

Detects outbound data sending attempts

skillaudit-obfuscationhigh

Obfuscated Content

Detects base64, hex, or URL shortener obfuscation

skillaudit-privilege-escalationcritical

Privilege Escalation Attempt

Detects privilege escalation attempts

skillaudit-prompt-injection-html-commentscritical

Hidden HTML Comment Instructions

Detects LLM instructions hidden in HTML comments

skillaudit-prompt-injectioncritical

Prompt Injection Pattern

Detects LLM system prompt override attempts

skillaudit-supply-chain-dynamicmedium

Dynamic External Reference

Detects dynamic external content pointers

skillaudit-supply-chain-packageshigh

Hallucinated Package Reference

Detects npm/pip package references for verification

Supply-chain verification

Package references are checked against public registries (HEAD request):