contract-review
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill reads external contract documents which represents a potential surface for indirect prompt injection attacks.
- Ingestion points: Legal contracts provided as PDF or text files are read in Step 1.
- Boundary markers: The instructions do not define specific delimiters or instructions to ignore embedded commands within the contract text.
- Capability inventory: The skill uses tools like search_all_summaries, search_notion, and search_emails to retrieve sensitive internal data for context.
- Sanitization: There is no mention of sanitizing or validating the input text before processing.
- [SAFE]: The skill instructions are transparent and follow a logical, professional workflow for contract analysis with no evidence of malicious intent or behavior.
Audit Metadata