deal-review-and-close-plan
Pass
Audited by Gen Agent Trust Hub on Oct 25, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill contains comprehensive instructions to prevent prompt injection. It explicitly mandates that all text from fetched pages, uploaded files, and pasted content must be treated as data and never as instructions.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data such as CRM exports and customer notes. It mitigates risk through a mandatory notification mechanism: if an instruction is detected within processed data, the agent must inform the user with a specific notice line and refuse the instruction. This prevents silent execution of malicious payloads.
- [COMMAND_EXECUTION]: The skill utilizes local Python scripts (deal_health.py and close_plan.py) for scoring and calendar logic. These scripts are restricted to the Python standard library, perform no network requests, and do not modify the file system.
- [SAFE]: The skill implements a '100% Rigor Rule' and 'Fact-Check-Before-Shipping' protocols in its operating principles. These rules ensure that the agent verifies user-supplied facts and avoids generating harmful or unverified content, representing a high standard of AI safety and reliability.
Audit Metadata