customer-support-verification
Pass
Audited by Gen Agent Trust Hub on Aug 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (customer emails and Gmail drafts) while having access to sensitive tools (Gmail, Stripe, Firebase, Git). This creates an attack surface for indirect prompt injection. * Ingestion points: Processes customer requests and Gmail thread/draft content as described in the 'Required Inputs' section of SKILL.md. * Boundary markers: The instructions do not define specific delimiters for untrusted content, though they provide verbal warnings. * Capability inventory: The skill interacts with Gmail (sending/archiving), Stripe/Firebase/Supabase (mutations/billing), and Git (staging/committing) as noted in the 'Verification Checklist' of SKILL.md. * Sanitization: Checklist item 5 provides a mitigation by instructing the agent to treat email links, attachments, and customer claims as untrusted.
- [COMMAND_EXECUTION]: The skill directs the agent to perform and verify actions involving Git repositories and administrative databases. * Evidence: Checklist items 12 (Commit discipline) and 6 (Mutation safety) in SKILL.md reference staging/committing files and performing mutations in Stripe, Firebase, or Supabase.
Audit Metadata