meta-cognition-parallel

Pass

Audited by Gen Agent Trust Hub on Mar 24, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill interpolates raw user input from $ARGUMENTS directly into prompts for sub-agents without robust sanitization or strict boundary markers.
  • Ingestion points: $ARGUMENTS in SKILL.md.
  • Boundary markers: Uses simple markdown headers (## User Query) which are insufficient to prevent injection.
  • Capability inventory: Launches Task sub-agents with general-purpose capabilities.
  • Sanitization: None detected.
  • [COMMAND_EXECUTION]: The skill attempts to access and read files using relative paths outside of its immediate directory (../../agents/). While these files appear to be legitimate instruction templates, the use of directory traversal sequences to access adjacent folders is a pattern that can be used for unauthorized file system discovery.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 24, 2026, 02:57 PM