customer-voice
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- Indirect Prompt Injection Surface: The skill processes information from external sources like email threads, transcripts, and documents identified in
SKILL.md. These are ingestion points for untrusted data. The skill implements boundary markers by instructing the agent to treat this as content only and never as instructions. It manages capabilities such as CRM and Gong reads and applies sanitization by suppressing embedded links and requiring user confirmation for content-originated actions. - Controlled Link Rendering: The skill prevents the rendering of links found within untrusted content, as seen in the rules of
SKILL.md. This proactive measure reduces the risk of unauthorized redirection or phishing by requiring the agent to use trusted system record identifiers instead.
Audit Metadata