diagnosing-bugs
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [SAFE]: The skill provides instructional content for debugging and does not contain any malicious code, obfuscated commands, or persistence mechanisms.
- [INDIRECT_PROMPT_INJECTION]: The skill interacts with untrusted data such as user bug reports, log dumps, and captured traces (SKILL.md). It mandates a 'Redact' phase where secrets and authentication headers are replaced with placeholders, effectively mitigating the risk of credential exposure through indirect injection.
- [DYNAMIC_EXECUTION]: The skill involves the creation of local harnesses, tests, and scripts for bug reproduction (Phase 1) and provides a template for human-in-the-loop scripts (scripts/hitl-loop.template.sh). These activities are standard for debugging and are performed within the local environment without remote code dependency.
Audit Metadata