The Agent Skills Directory

[SAFE]: The skill serves as a comprehensive educational guide for LangChain architecture. All provided code snippets are standard implementation templates for AI agents and Retrieval-Augmented Generation (RAG) pipelines.
[INDIRECT_PROMPT_INJECTION]: The skill provides instructions and code for ingesting external data through components like TextLoader and DirectoryLoader. This creates a theoretical attack surface common to all RAG and agentic systems, where instructions embedded in external files could attempt to influence the agent's behavior.
Ingestion points: loader = TextLoader('documents.txt'), loader = DirectoryLoader('./docs') in SKILL.md.
Boundary markers: The provided templates do not explicitly include boundary delimiters (like XML tags or markdown blocks) or specific instructions to ignore embedded commands in the source documents.
Capability inventory: The skill demonstrates initialize_agent with tools and RetrievalQA chains, which are capable of data retrieval and processing in SKILL.md.
Sanitization: No specific sanitization or filtering logic is provided in the educational examples for document content.
[COMMAND_EXECUTION]: The skill demonstrates how to configure autonomous agents using initialize_agent. While these are educational examples, agents configured with tools like serpapi or custom functions possess the capability to execute actions based on LLM decision-making.

langchain-architecture