The Agent Skills Directory

[SAFE]: The skill serves as a development guide for LangChain 1.0. Analysis of all scripts and documentation reveals no malicious patterns, command injections, or unauthorized data access.\n- [INDIRECT_PROMPT_INJECTION]: The skill manages untrusted data from AI models. 1. Ingestion points: AIMessage.content handled in SKILL.md and references/content-blocks.md. 2. Boundary markers: The skill encourages the use of AIMessage.text() and with_structured_output for reliable parsing. 3. Capability inventory: The skill metadata allows Read, Write, Edit, Bash, and Grep tools. 4. Sanitization: Logic relies on LangChain's internal normalization of provider-specific responses. No exploitable chains or malicious instructions were found.

langchain-model-inference