The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references standard, well-known Java dependencies (io.qdrant:client and dev.langchain4j) from the official Maven Central repository. It also provides standard Docker instructions to pull the official Qdrant image from Docker Hub. These represent normal development resources and do not pose a security risk.
[INDIRECT_PROMPT_INJECTION]: The skill provides patterns for building Retrieval-Augmented Generation (RAG) systems, which are inherently exposed to indirect prompt injection when processing untrusted external data.
Ingestion points: SearchController.search (query parameters) and RagController.ingestDocument (request body) in references/examples.md.
Boundary markers: The skill explicitly addresses this risk in the 'Constraints and Warnings' section of SKILL.md, instructing developers to 'Sanitize all document content before ingestion' and 'Apply content filtering on retrieved documents'.
Capability inventory: The skill is limited to vector database operations (upsert, query) and does not provide arbitrary command execution or system-level access.
Sanitization: While the simplified code examples do not implement full sanitization logic, the accompanying documentation provides clear guidance on its necessity for production systems.

qdrant