grilling
Pass
Audited by Gen Agent Trust Hub on Aug 19, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data in the form of user-submitted plans, decisions, or ideas.
- Ingestion points: The skill directly ingests user input regarding a "plan, decision, or idea" to build a design tree (SKILL.md).
- Boundary markers: The instructions do not define specific delimiters or instructions to ignore embedded commands within the user's input.
- Capability inventory: The skill explicitly instructs the agent to "dispatch a sub-agent to find it" whenever a fact from the environment (filesystem, tools, etc.) is needed (SKILL.md). This indicates high autonomy and access to system tools.
- Sanitization: There is no mention of sanitizing or validating the input before the sub-agent is dispatched to query the environment.
- Risk: An attacker could provide a plan containing malicious instructions that are processed when the sub-agent attempts to "find facts" from the environment.
Audit Metadata