starlight-investigate
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data such as software logs, reproduction steps, and system environment details to perform root-cause analysis.\n
- Ingestion points: Ingests logs, state transitions, and user-provided reproduction steps (SKILL.md).\n
- Boundary markers: Employs a workflow that strictly separates raw observations from hypotheses to prevent data content from influencing agent logic.\n
- Capability inventory: Uses standard diagnostic tools including repository search, test runners, and log viewers.\n
- Sanitization: Explicitly mandates the redaction of credentials and personal data before processing external diagnostic information.\n- [COMMAND_EXECUTION]: The skill involves executing software tests and diagnostic scripts to verify failures and fixes.\n
- Evidence: Workflow steps 5 and 7 mention running discriminating checks and regression tests. The tools section lists focused tests as a preferred utility.\n
- Context: Execution is scoped to diagnostic purposes and regression testing following user-authorized fixes.
Audit Metadata