inversion-exercise

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFENO_CODE
Full Analysis
  • [SAFE]: The skill consists entirely of markdown documentation and instructional content for a cognitive exercise. No executable scripts or system-level operations were found.
  • [NO_CODE]: No code blocks, shell commands, or external dependencies are present in the skill, precluding common attack vectors like command injection or remote code execution.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-defined assumptions but mitigates potential risks through the 'Anti-Patterns' section, which explicitly instructs the agent not to treat external content, logs, or tool responses as trusted instructions.
  • [DATA_EXFILTRATION]: There are no network operations (curl, wget, etc.) or file system read/write operations that could facilitate data exposure or exfiltration.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 03:51 AM
Security Audit — agent-trust-hub — inversion-exercise