campaign-sending
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external files that could contain malicious instructions designed to influence the agent.\n
- Ingestion points: The skill reads email content from
campaigns/{slug}/emails/tier{N}_emails.csvand campaign parameters fromcampaigns/{slug}/context.md. It also queries Salesforce data via a HubSpot MCP.\n - Boundary markers: The instructions do not define delimiters or provide specific guidance to the agent to ignore instructions that may be embedded within the CSV or Markdown content.\n
- Capability inventory: The skill has the authority to write logs to the filesystem and initiate network communications with external APIs (Gmail, Instantly, HeyReach, Nooks).\n
- Sanitization: No input validation, escaping, or sanitization processes are specified for the data retrieved from external sources before it is processed or transmitted.\n- [DATA_EXFILTRATION]: The skill is designed to transmit sensitive contact information and communication logs to external services.\n
- Evidence: Prospect names, email addresses, and company details are uploaded to Instantly, HeyReach, and Nooks APIs as part of the campaign workflow (file: SKILL.md).\n
- Context: While these data transfers are required for the skill's primary function of email and multi-channel outreach, they represent a structured movement of local data to third-party endpoints.
Audit Metadata