The Agent Skills Directory

[PROMPT_INJECTION]: The skill facilitates the ingestion of external document data into the agent's context, which is the standard architecture for RAG systems. This ingestion process creates a potential surface for indirect prompt injection. However, the skill provides mitigation through the use of structured prompt templates with clear boundary markers such as 'Context:' and 'Question:' to ground the agent's response behavior. Evidence: Ingestion points via 'DirectoryLoader' in 'sub-skills/4-reranking.md'; boundary markers present in 'sub-skills/contextual-prompt.md' and 'sub-skills/with-citations.md'; capability inventory includes standard retrieval and text generation; no explicit input sanitization logic is provided beyond these delimiters.
[EXTERNAL_DOWNLOADS]: The skill's instructions and code snippets reference and utilize official client libraries and APIs from well-known technology providers including OpenAI, Pinecone, and Weaviate. These external connections are necessary for generating embeddings and performing vector similarity searches, which are core functions of the described skill.

rag-implementation