grilling
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user answers to dynamically generate a 'design tree', which in turn triggers automated fact-finding via sub-agents.
- Ingestion points: User-provided answers to the frontier questions serve as the primary input (SKILL.md).
- Boundary markers: The instructions lack explicit delimiters or safety warnings to treat user answers as untrusted data.
- Capability inventory: The skill specifies the use of sub-agents to access the filesystem and tools to retrieve facts based on the evolving design tree (SKILL.md).
- Sanitization: No sanitization, validation, or escaping logic is defined for the user input before it influences the sub-agent's tool usage or filesystem access.
Audit Metadata