investigate-failing-tests
Pass
Audited by Gen Agent Trust Hub on Feb 20, 2026
Risk Level: SAFE
Full Analysis
- [SAFE] (SAFE): No malicious behavior, obfuscation, or safety bypass attempts were detected. The skill's logic is consistent with its stated diagnostic purpose.\n- [COMMAND_EXECUTION] (SAFE): The skill does not directly execute system commands. It delegates test execution and code modification to external skills, minimizing the risk of unauthorized command execution.\n- [DATA_EXFILTRATION] (SAFE): No network operations or credential exposures were found. File system access is restricted to reading local project configuration and context files (e.g., .webdriverio-skills/project-context).\n- [PROMPT_INJECTION] (SAFE): The skill focuses on reasoning and routing; it contains no instructions targeting agent constraints or attempting to extract system prompts.\n- [INDIRECT_PROMPT_INJECTION] (SAFE): While the skill ingests user input and error logs (ingestion points: failing spec path, error message, stack trace), it includes boundary-like behaviors such as requiring explicit user confirmation before acting on hypotheses, which reduces the surface for indirect injection (capability inventory: delegating to other skills).
Audit Metadata