meta-cognitive-reasoning

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFE
Full Analysis
  • [Prompt Injection] (SAFE): The document contains examples of analysis using markers like 'CRITICAL' and 'BLOCKER', but these are used as templates for self-correction and reasoning, not as instructions to override agent behavior.
  • [Command Execution] (SAFE): Shell commands such as 'ls', 'grep', and 'mkdir' are presented as hypothetical evidence-gathering steps or CI/CD observations and are not executed by the skill itself.
  • [Data Exposure] (SAFE): No sensitive file paths, environment variables, or hardcoded credentials were detected.
  • [External Downloads] (SAFE): References to package versions like 'certifi==2025.10.5' are used for temporal reasoning examples and do not involve remote code execution or untrusted source downloads.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 05:48 PM