grilling
Pass
Audited by Gen Agent Trust Hub on Aug 20, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructions allow the agent to dispatch sub-agents to retrieve facts from the environment, such as the filesystem or other tools, based on user-provided context. This creates a surface area where external data could influence agent logic if that data contains embedded instructions.
- Ingestion points: Plan details and environmental facts retrieved from tools or the filesystem as described in SKILL.md.
- Boundary markers: The instructions lack explicit delimiters or specific 'ignore' directives to separate retrieved data from core system logic.
- Capability inventory: The agent is empowered to access the filesystem, use local tools, and spawn sub-agents to process data.
- Sanitization: No specific sanitization or validation methods are described for handling potentially untrusted content retrieved from the environment.
Audit Metadata