vendor-ai-review
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- Indirect Prompt Injection Surface: The skill is designed to ingest and analyze external, untrusted data in the form of vendor contracts, terms of service, and AI addendums. This creates a potential surface where hidden or malicious instructions within these documents could attempt to influence the agent's analysis or behavior.
- Ingestion points: The skill reads vendor-provided PDFs and text documents (e.g.,
openai-enterprise-agreement.pdf) and local configuration files (CLAUDE.md,matter.md) to perform its review. - Boundary markers: The instructions do not explicitly define boundary markers or instruct the agent to ignore potentially adversarial instructions embedded within the contract text.
- Capability inventory: The skill utilizes local file system read and write operations within a specific plugin directory (
~/.claude/plugins/config/claude-for-legal/). It does not demonstrate network access, arbitrary command execution, or other high-privilege capabilities. - Sanitization: There is no evidence of sanitization or filtering applied to the external document content before processing.
Audit Metadata