grilling
Pass
Audited by Gen Agent Trust Hub on Aug 12, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes unconstrained user input to construct a 'design tree' and determine the frontier of questions.
- Ingestion points: User answers provided during the interview process defined in SKILL.md.
- Boundary markers: None. The instructions lack delimiters or warnings to ignore embedded instructions within user-provided data.
- Capability inventory: The skill explicitly authorizes the agent to dispatch sub-agents and use tools to access the filesystem and environment to find 'facts' (SKILL.md).
- Sanitization: None. There is no evidence of input validation, output sanitization, or filtering for data retrieved from the environment.
- [COMMAND_EXECUTION]: The instructions direct the agent to autonomously use tools and explore the filesystem when 'facts' are needed. This autonomous behavior, when triggered by user-influenced conversation paths, could lead to unexpected tool execution or data exposure if not properly constrained.
Audit Metadata