slowmist-agent-security-framework

Pass

Audited by Gen Agent Trust Hub on May 18, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill documentation includes examples of injection strings (e.g., "ignore all previous instructions") within code blocks for a detection function. These are included for educational and testing purposes to demonstrate malicious behavior for identification, and are not intended for execution by the agent loading the skill.
  • [OBFUSCATION]: A base64-encoded URL is provided within a code example to illustrate how to detect hidden network endpoints. The decoded URL (https://evil.com/steal) is part of a theoretical malicious scenario used to teach the agent rejection criteria.
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions to download security tools and pattern databases from the official GitHub organization of SlowMist, a recognized security firm.
  • [COMMAND_EXECUTION]: The documentation describes using standard CLI utilities like git and curl for the purpose of auditing repositories and analyzing network responses.
Audit Metadata
Risk Level
SAFE
Analyzed
May 18, 2026, 05:03 PM
Security Audit — agent-trust-hub — slowmist-agent-security-framework