simulator-effect

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFE
Full Analysis
  • SAFE (SAFE): Extensive review of the skill files confirms the absence of malicious patterns, hardcoded credentials, or dangerous commands. The skill serves as a conceptual framework for MOOLLM design.
  • METADATA_POISONING (LOW): The file skill-snitch-report.md contains a self-referential verdict ('THE MOST POWERFUL INSIGHT. APPROVE'). This attempt to influence the auditor's judgment was detected and ignored in accordance with the 'assume-malicious' posture.
  • INDIRECT_PROMPT_INJECTION (LOW): The 'YAML Jazz' method describes treating data-level comments as instructions ('seeds'). This represents a potential vulnerability surface.
  • Ingestion points: The skill facilitates reading YAML files where comments are intended to influence LLM behavior.
  • Boundary markers: No explicit delimiters or 'ignore embedded instructions' warnings are provided for these seeds.
  • Capability inventory: The skill is restricted to read_file operations.
  • Sanitization: No sanitization logic is present, as the methodology relies on the LLM interpreting the 'seeds' directly.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:48 PM