systematic-debugging
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill workflow relies on processing external data such as error messages, stack traces, and component logs. This represents an indirect prompt injection surface where malicious instructions embedded in logs could influence agent behavior. Ingestion points include SKILL.md (Phase 1) and scripts/hitl-loop.template.sh (capture ERROR_MSG). The instructions do not specify using delimiters or 'ignore instructions' warnings for these inputs, and no sanitization of the gathered log or error data is performed.
- [COMMAND_EXECUTION]: The skill provides a bisection utility for identifying problematic tests. The script scripts/find-polluter.sh executes npm test on files matching a user-defined pattern.
- [DATA_EXFILTRATION]: The skill provides diagnostic examples that access sensitive system configuration information. SKILL.md (Phase 1, Step 4) includes bash examples using security list-keychains and security find-identity -v which expose keychain file paths and signing identity metadata for troubleshooting purposes.
Audit Metadata