systematic-debugging
Pass
Audited by Gen Agent Trust Hub on Mar 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [PROMPT_INJECTION]: The skill defines a rigid operational framework termed 'The Iron Law' and four mandatory phases which act as behavioral overrides for the agent's default problem-solving logic.
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface due to the processing of untrusted logs and code fragments. 1. Ingestion points: Error messages, stack traces, and git diffs (SKILL.md). 2. Boundary markers: Absent. 3. Capability inventory: Local subprocess execution and file system access. 4. Sanitization: Absent.
- [COMMAND_EXECUTION]: The 'find-polluter.sh' script automates the execution of 'npm test' on project files, which constitutes local code execution.
- [DATA_EXFILTRATION]: The skill suggests diagnostic steps that reveal sensitive system state, such as environment variables ('env | grep') and keychain identities ('security list-keychains'), which can expose secrets if recorded in logs.
Audit Metadata