vtex-io-application-performance
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill documents a capability surface that ingests data from external VTEX platform APIs and uses it to perform high-privilege operations.
- Ingestion points: External data enters the agent context through
ctx.clients(e.g.,search.getSkuById,checkout.orderForm,getSessions) andAppSettingsloading. - Boundary markers: The skill explicitly mandates the use of
accountandworkspaceidentifiers from the VTEX context as mandatory tenant boundary markers for all in-memory caches to prevent cross-tenant data access. - Capability inventory: The skill instructions cover the use of
ctx.clientsfor network operations (including payment authorization calls) andvbasefor persistent data storage. - Sanitization: The skill mandates the redaction of sensitive data in logs and provides strict rules for isolating tenant-specific data within shared module-level memory.
- [DATA_EXFILTRATION]: The skill provides robust guidelines to prevent unauthorized data exposure. It dictates that PII, tokens, and authorization-sensitive data must never be stored in global caches without explicit tenant-based isolation keys. It also provides guidance on redacting sensitive data when using platform logging tools.
- [SAFE]: No malicious patterns or obfuscation techniques were detected. The skill promotes security best practices, such as awaiting critical persistence writes in financial paths to ensure idempotency and configuring explicit timeouts for all outbound network requests.
Audit Metadata