Managed Deep Agents - Skills
LangChain's managed deep agents support on-demand skill loading via progressive disclosure
“Skills are basically context and scripts that your agent can run and read dynamically as it's running.”
LangChain released a tutorial on adding skills to managed deep agents, where skills are modular folders of context and scripts loaded on demand rather than injected into every prompt. The mechanism, called progressive disclosure, shows the agent only a skill name and description upfront, letting it choose to load full content when needed. This addresses a practical token-efficiency and context-bloat problem in long-running agents, but represents incremental tooling rather than a major shift.