campaign-feedback
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources including email replies, LinkedIn messages, and CRM records. This creates a potential surface for indirect prompt injection where adversarial content in prospect communications could attempt to influence the agent's analysis or knowledge graph updates.
- Ingestion points: Step 1 (Import Results) reads from send logs, reply data, and CRM platforms.
- Boundary markers: None explicitly defined for untrusted reply content.
- Capability inventory: File writing to
report.mdand API calls to a knowledge graph client. - Sanitization: No explicit sanitization or filtering of external reply content is specified before processing.
Audit Metadata