systematic-debugging
Pass
Audited by Gen Agent Trust Hub on Jun 30, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill inherently involves processing untrusted data, such as error logs, stack traces, and user-provided reproduction inputs, which is necessary for debugging tasks.
- Ingestion points: The skill instructs the agent to read external logs, error context, and input data in Phase 1 (Gather Symptoms) and Phase 2 (Reproduce).
- Boundary markers: The instructions do not explicitly mandate the use of delimiters or 'ignore embedded instructions' warnings when analyzing this external data.
- Capability inventory: The process involves file system operations (reading logs, writing debug files) and shell command execution (e.g.,
cat,tail). - Sanitization: No specific sanitization or validation of the ingested logs or error messages is defined in the instructions.
Audit Metadata