edi-engineer
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to process external B2B data (EDI documents), which creates an attack surface for indirect prompt injection.\n
- Ingestion points: Inbound X12 and EDIFACT messages (e.g., 850 Purchase Orders, ORDERS) received through transport protocols like AS2, SFTP, and VAN as described in
references/edi_engineer_scope.mdandreferences/transport_and_partner_onboarding.md.\n - Boundary markers: The guidance requires implementing multiple validation layers (syntax, partner-specific implementation guides, and business rules) to ensure data conforms to expected schemas before processing (
references/validation_and_acknowledgments.md).\n - Capability inventory: The agent is guided to transform EDI data into canonical models, post these to internal systems (ERP, WMS, OMS) via APIs, and manage document persistence in object stores (
references/testing_monitoring_and_reconciliation.md).\n - Sanitization: The instructions mandate a "fail closed" approach to validation, requiring all malformed data to be quarantined. Furthermore, there is a strict requirement to redact Protected Health Information (PHI) from all logs and system outputs.
Audit Metadata