contract-review
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content in the form of contract documents which constitutes an indirect prompt injection surface.
- Ingestion points: Contract documents (PDF, text) or counterparty names provided via the main command argument.
- Boundary markers: The skill incorporates a disclaimer block and metadata headers but does not specify strict structural delimiters (e.g., XML tags) to isolate the untrusted contract text from the agent's instructions.
- Capability inventory: The skill can read local files and write its analysis to a local markdown file but has no capabilities for network operations or shell command execution.
- Sanitization: Relies on referenced compliance with internal legal guardrails and a predefined analytical pass to maintain focus on clause-by-clause triage rather than executing embedded instructions.
Audit Metadata