grill-me
Pass
Audited by Gen Agent Trust Hub on Mar 13, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns, obfuscation, or exfiltration attempts were detected. The skill focuses on facilitate shared understanding through dialogue.
- [PROMPT_INJECTION]: The directive to interview the user relentlessly is a behavioral instruction consistent with the skill's purpose and does not attempt to override system safety guidelines.
- [DATA_EXPOSURE]: The agent is instructed to explore the codebase to answer questions, which implies read access to project files but is standard for design and code review tasks.
- [INDIRECT_PROMPT_INJECTION]: The skill creates a surface for indirect injection by ingesting data from the codebase. Ingestion points: local codebase files; Boundary markers: absent; Capability inventory: codebase file read access; Sanitization: absent.
Audit Metadata