The Agent Skills Directory

[SAFE]: The skill includes explicit security warnings in reference/cn-models/README.md, instructing developers never to read or write sensitive user configuration files such as .env or credentials.json.
[SAFE]: The template/chat_model.py follows best practices for secret management by using secret_from_env and from_env to handle API keys and base URLs rather than hardcoding credentials.
[SAFE]: Instructions in reference/deepagents.md regarding filesystem access and shell execution include clear warnings about security implications, such as the risks of path traversal and the importance of using sandboxes for shell commands in production environments.
[SAFE]: The external libraries and tools mentioned, including ecosystem packages like langchain-openai and vendor-provided tools like contextseek and langchain-dev-utils from ob-labs, are consistent with the skill's stated purpose of providing engineering fixes and integrations for the LangChain ecosystem.

langchain-dev-guide