debug-like-expert
Pass
Audited by Gen Agent Trust Hub on Mar 13, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is instructional and does not contain executable code, hidden commands, or external package dependencies. It explicitly enforces a 'read-only' mindset and a mandatory 'decision gate' for all recommendations, which are significant security best practices.
- [PROMPT_INJECTION]: Analysis of indirect prompt injection surface: 1. Ingestion points: Reads source code and error messages from the local environment and web search results (SKILL.md). 2. Boundary markers: Absent; there are no specific instructions to wrap or isolate analyzed content. 3. Capability inventory: MCP servers, web search, and file read operations (SKILL.md). 4. Sanitization: Absent. Mitigation: The skill includes a high-severity instruction to never modify code and to present findings only through a mandatory human-in-the-loop decision gate, effectively neutralizing potential injection attacks aiming for code modification.
Audit Metadata