analyzing-test-failures
Pass
Audited by Gen Agent Trust Hub on Jun 23, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is defined as 'Pure reasoning' and explicitly prohibits system calls or code authoring, which prevents a wide range of execution-related attacks.
- [DATA_EXPOSURE]: The skill includes instructions to avoid exposing raw technical data, such as stack traces and internal error codes, to the user. This practice reduces the risk of disclosing sensitive system architecture information during failure analysis.
- [PROMPT_INJECTION]: The skill processes untrusted external data (JSON failure payloads). While this creates a theoretical surface for indirect prompt injection, the risk is negligible because the skill lacks the tools or permissions to perform actions that could be exploited (e.g., file writes or network requests).
Audit Metadata