grilling
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest user-supplied plans, decisions, or ideas and proactively verify environmental facts (using tools or the filesystem) via sub-agents. This pattern creates a surface for indirect prompt injection where malicious instructions embedded in a user's plan could be interpreted as directives for the sub-agent during fact-finding operations.
- Ingestion points: User input regarding plans, decisions, or ideas processed during the 'grilling' session (SKILL.md).
- Boundary markers: The instructions lack explicit delimiters or warnings to treat user-provided data as untrusted content when performing environment lookups.
- Capability inventory: The skill explicitly directs the agent to access the filesystem and environment tools through sub-agent dispatch (SKILL.md).
- Sanitization: There is no requirement for input validation or sanitization before the agent executes tool-based lookups.
Audit Metadata