ai-native-audit-release

Pass

Audited by Gen Agent Trust Hub on Jun 27, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: No override markers, bypass instructions, or system prompt extraction patterns were detected. The instructions are focused on guiding the AI through a structured auditing methodology.
  • [DATA_EXFILTRATION]: No network operations (curl, wget, etc.) or access to sensitive local file paths (e.g., credentials, .env files) were found. The skill operates entirely within the chat context.
  • [CREDENTIALS_UNSAFE]: No hardcoded API keys, tokens, or private keys are present in the skill documentation or metadata.
  • [REMOTE_CODE_EXECUTION]: No remote script downloads or dynamic code execution patterns (eval, exec) were detected. The skill does not install any external packages or dependencies.
  • [COMMAND_EXECUTION]: The skill does not utilize any shell commands, subprocess calls, or privilege escalation techniques.
  • [OBFUSCATION]: No Base64, zero-width characters, homoglyphs, or other encoding techniques were found. All content is clear and human-readable.
  • [SAFE]: The skill serves as a template and decision-making guide for evaluating other systems' safety and reliability. It mentions security auditing (including checking for prompt injection in target systems) as a best practice, which is a defensive measure.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 27, 2026, 07:51 AM
Security Audit — agent-trust-hub — ai-native-audit-release