The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents a Retrieval-Augmented Generation (RAG) pattern, which is inherently susceptible to indirect prompt injection if the ingested documents contain malicious instructions.
Ingestion points: Data enters the system via the ingestDocument function in SKILL.md.
Boundary markers: The code uses --- delimiters between context chunks and a clear Context: header in the system prompt.
Capability inventory: The documented pattern includes vector database operations and LLM text generation; it does not include dangerous OS-level capabilities.
Sanitization: The examples do not demonstrate explicit escaping or sanitization of retrieved context before prompt interpolation.
[EXTERNAL_DOWNLOADS]: The skill references standard, well-known libraries for AI application development.
References @ai-sdk/openai and ai packages (Vercel AI SDK).

rag-patterns