The Agent Skills Directory

[PROMPT_INJECTION]: The skill documents RAG implementation patterns which involve processing external, untrusted content. This represents an indirect prompt injection surface. 1. Ingestion points: Context is retrieved from vector stores in several code examples throughout SKILL.md. 2. Boundary markers: The prompt templates use standard delimiters but lack specific instructions to the model to ignore embedded commands within the context. 3. Capability inventory: The skill provides logic for document retrieval and LLM answer generation. 4. Sanitization: The examples do not demonstrate sanitization or filtering of the retrieved document content.

rag-implementation