agentsop-hybrid-retrieval
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill outlines a retrieval architecture that ingests and processes untrusted user queries, creating an attack surface for indirect prompt injection.
- Ingestion points:
SKILL.mddescribes the ingestion of user-generated lexical and semantic queries into theQueryFusionRetrieverpipeline. - Boundary markers: The provided instructions do not specify the use of delimiters or boundary markers to isolate untrusted input from the agent's instructions.
- Capability inventory: The skill provides logic for Python-based retrieval and rank fusion using established frameworks like LlamaIndex and LangChain.
- Sanitization: No input validation or sanitization strategies for user queries are described.
- [EXTERNAL_DOWNLOADS]: The skill references technical documentation and informational blog posts from well-known service providers in the AI ecosystem.
- Evidence: Informational links are provided to the official domains and repositories of LlamaIndex, LangChain, Qdrant, Weaviate, and Pinecone.
Audit Metadata