Field noteOct 23, 20261 min read1 source

Skills - Docs by LangChain

In one minute

LangChain’s Deep Agents documentation describes versioned controls for loading and operating agent skills.

What changed: Same event as Revamping Skills in Deep Agents

A skill is a directory with a SKILL.md file containing YAML frontmatter and instructions, with optional scripts, references, templates, and other resources. Deep Agents follows the Agent Skills specification.

Tool binding is documented for deepagents>=0.7.22. A skill can list space-separated tool names in metadata.include_tools; those tools remain gated until the agent reads that skill’s SKILL.md. Tools are attached through SkillsMiddleware, and a resolver can map names to tools at runtime, including per-user tools.

Pinned skills are documented for >=0.7.23: names supplied in pinned_skills cause the corresponding SKILL.md content, excluding frontmatter, to be appended as tagged messages before the next model call, and make that skill’s tools available. Reload support is documented for >=0.7.16: setting skills_metadata to None in run input or through update_state rescans skills on the next run; a changed set invalidates the prompt cache.

The documentation does not provide a release date, benchmark results, or evidence that these controls improve skill selection or task success rates.

Evidence / sources

  1. 1docs.langchain.com/oss/python/deepagents/skills
All Field Notes

Next reads

Read next

Open the graph
  1. 01topicSkillsExplore the skills topic hub.
  2. 02archiveField NotesOpen the latest editorial analysis.
  3. 03source ledgerSource LedgerInspect the public source evidence graph.
Who read this page?Machine requests, hidden until opened

Loading the privacy-safe route aggregate...

Open the JSON contract