grilling
Pass
Audited by Gen Agent Trust Hub on Nov 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill seeds its design tree by reading external files including project plans and decision logs. If these files contain adversarial instructions, the agent could be manipulated during the 'grilling' process. * Ingestion points: Reads project plans and decision records from the local environment (SKILL.md). * Boundary markers: No explicit markers or instructions to ignore embedded commands are present. * Capability inventory: Authorized to read/write project files and use tools/sub-agents for environmental discovery. * Sanitization: No content validation or sanitization is described for the ingested data.
- [COMMAND_EXECUTION]: The instructions direct the agent to autonomously 'dispatch a sub-agent' to find facts from the filesystem and tools without user prompting. This grants the agent autonomy to explore the local environment beyond the immediate chat context.
Audit Metadata