grimoire-interrogator

Pass

Audited by Gen Agent Trust Hub on Jul 22, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it dynamically updates its decision-making logic based on untrusted user answers.\n
  • Ingestion points: User answers collected in Step 4 are processed in Step 5 to generate child nodes and 'reshape' the design tree.\n
  • Boundary markers: The instructions do not specify delimiters or safety markers to isolate user input from the agent's internal reasoning logic.\n
  • Capability inventory: The skill is capable of dispatching sub-agents to access file contents, tool outputs, and environment data to perform fact-checks.\n
  • Sanitization: No input validation or sanitization is defined for user-provided answers before they are used to influence the agent's logic flow.\n- [COMMAND_EXECUTION]: The skill instructs the agent to dispatch sub-agents to perform system-level information gathering.\n
  • Evidence: Step 2 of the 'Loop' section directs the agent to 'dispatch a sub-agent' for 'file contents, tool output, or environment data' to resolve frontier questions.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 22, 2026, 11:32 AM