diagnose
Pass
Audited by Gen Agent Trust Hub on May 4, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill defines a standard debugging methodology (Reproduce -> Minimise -> Hypothesise -> Instrument -> Fix) without introducing any malicious code, prompt injections, or unauthorized data access.
- [COMMAND_EXECUTION]: While the skill instructions encourage the agent to execute commands (such as running tests, using curl, or executing a bisection loop), these are part of the intended functionality for debugging and do not involve untrusted remote code execution or privilege escalation.
- [DATA_EXFILTRATION]: No evidence of sensitive data exposure or exfiltration. The included script
scripts/hitl-loop.template.shis a benign helper for collecting human input during debugging.
Audit Metadata