auto-triage
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data including triage-verifier proof packets and agent-focused PoC artifacts.
- Ingestion points: SKILL.md Step 1 loads external proof packets and artifacts.
- Boundary markers: No specific delimiters or safety warnings are defined for the ingested data.
- Capability inventory: The skill is capable of executing shell commands and HTTP requests in Step 7 to verify the reproduction flow.
- Sanitization: The skill redacts secrets and PII but lacks explicit sanitization for natural language instructions within the evidence.
- [DYNAMIC_EXECUTION]: The skill is instructed to test generated reproduction steps.
- Evidence: SKILL.md Step 7 requires the agent to run the human flow (shell commands and network requests) to catch errors, which involves execution of dynamically generated content.
Audit Metadata