agentsop-idempotent-ingestion
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a standard ingestion workflow that processes external documents, creating a vulnerability surface for indirect prompt injection attacks.
- Ingestion points: Documents are ingested via
SimpleDirectoryReaderor generic data loaders as described inSKILL.md(Stage 2) andreferences/R2-cross-framework.md. - Boundary markers: The instructions do not recommend specific boundary markers or "ignore embedded instructions" warnings for the ingested document content.
- Capability inventory: The ingestion pipeline utilizes embedding models (network access) and persists data to vector stores and document stores (file/network access).
- Sanitization: The SOP focuses on hash-based deduplication but does not specify procedures for sanitizing or filtering document content prior to embedding.
Audit Metadata