handle-checkout-flow
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill addresses the processing of untrusted data such as shipping addresses and coupon codes. It mitigates potential risks by mandating server-side recalculation of all monetary amounts and requiring third-party API validation for addresses before further processing. 1. Ingestion points: User-provided shipping addresses, discount codes, and client-side pricing data (SKILL.md). 2. Boundary markers: The instructions explicitly require the agent to discard client-submitted monetary amounts and use server-side logic. 3. Capability inventory: Tools include fetch for network requests and file system modification tools. 4. Sanitization: Employs mandatory external address validation (e.g., Google Maps, Loqate) and atomic server-side validation for coupons.
- [COMMAND_EXECUTION]: The upgrade plan includes steps to run internal validation scripts (core/scripts/validate-all.py) to ensure adherence to standards. These operations are scoped to the local development environment.
- [SAFE]: The skill incorporates high-security requirements including PCI DSS v4.0.1 (Requirement 6.4.3 and 11.6.1), Subresource Integrity (SRI) hashes, and cryptographically signed HMAC tokens for session management.
Audit Metadata