plugin-architecture-patterns
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines traits and patterns for document extractors that process untrusted external content, creating a potential surface for indirect prompt injection if documents contain malicious instructions.
- Ingestion points: The
extract_contentmethod inInternalDocumentExtractor(SKILL.md) accepts raw bytes (&[u8]) and MIME type strings as input. - Boundary markers: The provided code templates and documentation do not include specific delimiters or instructions to ignore embedded commands within the processed data.
- Capability inventory: The architecture allows extractors to populate InternalDocument and ExtractedDocument structures which are then used by the agent for further reasoning or task execution.
- Sanitization: There is no documentation regarding the sanitization, filtering, or escaping of data extracted from documents before it is presented to the LLM.
Audit Metadata