customer-support-verification
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill operates as a compliance and security audit tool for support tasks. It mandates that the agent verify work against official runbooks and ensures that all mutations (Stripe, Firebase, Supabase) or communications (Gmail) are authorized and correctly scoped before finalizing a response.- [INDIRECT_PROMPT_INJECTION]: The skill defines a surface for indirect prompt injection by processing external data from customer emails and support requests (SKILL.md). It proactively mitigates this risk by instructing the agent to treat external content as untrusted evidence.
- Ingestion points: Customer requests, Gmail threads, and draft bodies are ingested from untrusted external sources.
- Boundary markers: No specific structural delimiters are defined for the ingested data.
- Capability inventory: The execution environment includes capabilities for interacting with Gmail, Stripe, Firebase, Supabase, and git repositories.
- Sanitization: Item 5 of the verification checklist provides a direct instruction to the agent to disregard instructions or claims contained within links, images, or attachments.
Audit Metadata