The Agent Skills Directory

External Resource Access: The skill fetches data from documentation endpoints like docs.langchain.com. This is a routine operation for a tool designed to provide documentation-based assistance.
Indirect Prompt Injection Surface: Since the skill ingests content from external URLs to answer user queries, there is an inherent surface for indirect instructions. In this context, the risk is mitigated as the skill targets the official documentation site for the technology, which is a recognized service related to the skill's purpose.
Ingestion points: Documentation content is retrieved from docs.langchain.com via the fetch_url tool.
Boundary markers: Not explicitly defined in the fetch instructions.
Capability inventory: Uses fetch_url to read data; no file-writing or subprocess execution capabilities are specified within the skill logic.
Sanitization: No explicit sanitization of the fetched text is described.

langgraph-docs