risk-analysis
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external documents (e.g., proposals, term sheets, EULAs), which introduces a potential surface for indirect prompt injection attacks.
- Evidence Chain:
- Ingestion points: Source documents are identified as primary inputs in the 'Inputs Required' table and analyzed in 'Step 2'.
- Boundary markers: The skill mandates a specific 'Disclaimer + UPL Guardrail Block' at the top of its output and requires an explicit 'Jurisdiction Assumed' field to frame the AI's context.
- Capability inventory: The skill is limited to text analysis and file generation; it lacks network access, shell execution, or high-privilege capabilities.
- Sanitization: Mitigation is handled via a 'Reviewer Checklist' that forces human verification of material facts against source documents before findings are accepted.
Audit Metadata