email-search
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external contact files, which presents a surface for indirect prompt injection if the source data contains malicious instructions targeted at downstream consumers.
- Ingestion points: Processes contact information from JSON files in
campaigns/{slug}/contacts/*.jsonandcampaigns/{slug}/segmentation.json. - Boundary markers: Not explicitly defined in the workflow instructions.
- Capability inventory: Enrichment via HubSpot, ZoomInfo, and Clay (MCP); email verification via EVA; file writing for updated JSON and campaign CSVs.
- Sanitization: No specific sanitization or validation logic is detailed for the incoming contact fields beyond syntax checking for emails.
- [SAFE]: The skill uses established data providers and cloud services for its operations. There is no evidence of hardcoded credentials, unauthorized network exfiltration, or attempts to bypass agent security constraints.
Audit Metadata