debug-triage
Pass
Audited by Gen Agent Trust Hub on Jun 14, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data such as terminal transcripts and stack traces.
- Ingestion points: Terminal transcripts, stack traces, and PowerShell logs pasted by the user.
- Boundary markers: The instructions lack explicit delimiters or warnings to treat the pasted content as untrusted data.
- Capability inventory: The skill uses Read, Grep, Glob, and specific git subcommands (log, blame, diff) to investigate errors.
- Sanitization: No input validation or sanitization is performed on the pasted error logs.
- [DATA_EXFILTRATION]: The instructions direct the agent to "check the environment" (including environment variables and potentially configuration files like .env) when triaging errors. While standard for debugging environmental issues, this could lead to the exposure of sensitive configuration data to the agent context.
Audit Metadata