debug
Pass
Audited by Gen Agent Trust Hub on Oct 19, 2026
Risk Level: SAFE
Full Analysis
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill includes a 'Redact' section that mandates the removal of secrets and authentication headers from captured artifacts and logs before they are shown or processed. It also promotes the use of environment variables to keep credentials out of the command output.
- [INDIRECT_PROMPT_INJECTION]: Phase 2 contains a security warning identifying error output, stack traces, and CI logs as untrusted data. It explicitly forbids the agent from executing commands or visiting URLs found within such data, which is a key defense against indirect injection through logs.
- [DYNAMIC_EXECUTION]: The skill instructs the agent to create deterministic feedback loops using local scripts (e.g., Python, Node.js, Shell). These scripts are intended for local diagnostic purposes within the developer's environment to verify fixes for bugs and regressions.
- [COMMAND_EXECUTION]: The skill utilizes standard development tools like
curl,git bisect, and CLI invocations to diagnose issues. These are standard capabilities for an engineering-focused agent and are used within a structured debugging framework.
Audit Metadata