analyze-account-signals
Pass
Audited by Gen Agent Trust Hub on Jul 15, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [Indirect Prompt Injection Surface]: The skill processes data from external messaging, meeting notes, and document stores. This ingestion of untrusted content creates a surface for indirect instructions to influence the agent. The skill mitigates this by requiring explicit user approval before any automated actions are performed. Evidence chain: (1) Ingestion points: external_messaging, document_store, meeting_notes. (2) Boundary markers: Absent. (3) Capability inventory: drafting messages, posting/sharing/writing to CRM. (4) Sanitization: Absent.
- [Sensitive Data Access]: The skill is configured to access sensitive information such as CRM records and internal communications. This access is consistent with its purpose and is managed through the sales:user-context preflight script.
- [Human-in-the-loop Oversight]: The skill mandates that all actions like posting, sharing, or writing to CRM must follow explicit user approval, providing a robust security control against accidental or malicious automated actions.
Audit Metadata