chargeback-management-prevention
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructions involve processing untrusted data such as customer communications and dispute reasons from payment platforms like Shopify, Stripe, and PayPal. \n- Ingestion points: SKILL.md (Step 3) guides the agent to read external dispute data from payment dashboards. \n- Capability inventory: The agent is instructed to interact with payment platform APIs and user interfaces to submit evidence packages. \n- Boundary markers: There are no explicit instructions to delimit or ignore instructions that might be embedded in untrusted customer communications. \n- Sanitization: The skill lacks guidance on sanitizing or escaping customer-provided text before it is processed by the agent or submitted back to the payment network.
Audit Metadata