demo-confirm-and-handoff
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFECREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill exposes a hardcoded roster of internal SDR team members that includes sensitive system identifiers such as work emails, HubSpot owner IDs, and Nooks user IDs (e.g., 'UdhD9yeRfOXij75iD0Oy9F1b25P2').
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted discovery and account information from external platforms to draft emails and internal handoff packages. 1. Ingestion points: Contextual data retrieved via 'hubspot_get_contact', 'hubspot_get_deal', and Nooks session data. 2. Boundary markers: The skill does not employ explicit delimiters or 'ignore instructions' warnings when processing this external content. 3. Capability inventory: The skill can perform 'create_draft' in Gmail, 'create_event' in Calendar, and build 'ae_handoff' packages. 4. Sanitization: There is no evidence of sanitization to filter or escape malicious instructions embedded in the ingested discovery data.
Audit Metadata