The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides instructions and patterns for building Retrieval-Augmented Generation (RAG) applications using LlamaIndex. These applications, by design, ingest untrusted data from external sources (PDFs, text files, markdown) and interpolate them into LLM prompts. While this is the intended purpose of the skill, it creates an attack surface where malicious instructions embedded in documents could influence the agent's behavior.
Ingestion points: SimpleDirectoryReader and PDFReader are used to load data from the ./data directory (SKILL.md).
Boundary markers: The provided examples do not demonstrate the use of explicit delimiters or instructions to ignore embedded commands when processing retrieved text.
Capability inventory: The skill demonstrates building agents (ReActAgent) that can execute tools (QueryEngineTool) and perform network/file operations based on LLM outputs.
Sanitization: There is no mention of sanitizing or filtering retrieved content before it is passed to the LLM for response generation.

llamaindex-development