grilling
Pass
Audited by Gen Agent Trust Hub on Jul 16, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is designed for interactive questioning and stress-testing user ideas without introducing malicious scripts, obfuscation, or persistence mechanisms.
- [DATA_EXFILTRATION]: The skill instructs the agent to look up facts in the environment (filesystem or tools). This is a standard context-gathering capability for agentic skills and is not directed at sensitive paths or unauthorized external exfiltration.
- [PROMPT_INJECTION]: The instructions establish a specific role and conversational workflow without attempting to bypass safety protocols, extract system prompts, or override agent constraints.
- [INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it processes untrusted user data and has tool access. Evidence: 1. Ingestion points: User plans and decisions (SKILL.md). 2. Boundary markers: Absent. 3. Capability inventory: Filesystem and tool access (SKILL.md). 4. Sanitization: Absent. The risk is considered low given the primary function is conversational questioning.
Audit Metadata