liveBook · Manning

Skill activation

As explained in Build a Multi-Agent System (from Scratch)

Skill activation is the process of loading a skill's procedural workflow into the LLM's context so the workflow can be executed.

Skill activation loads a selected skill’s procedural workflow into the backbone LLM’s context. Before activation, the LLM has only the skill’s name and description; activation discloses the main body and information for locating additional resources. section 6.4

A tool-based activation flow makes the disclosure explicit:

  • The agent requests activation through UseSkillTool, supplying a valid skill name. section 6.4.1
  • The tool returns XML-structured content containing the skill’s main body and any additional resources. section 6.4.1
  • The agent can then execute the disclosed workflow to perform the task. section 6.1.1

UseSkillTool is added to the tools available during task execution. Users can also explicitly invoke a skill through the framework’s run_with_skill() method. section 6.4.1 section 6.6

How Build a Multi-Agent System (from Scratch) covers it

How deep it goes
Implements activation through a dedicated tool that returns XML containing the skill workflow and available resources.
Examples
Demonstrates activation as a UseSkillTool request carrying a skill name and returning the skill content as XML.
What it assumes
Assumes discovery has exposed a skill’s name and description, while deferring its procedural workflow until activation.
What you can do after
Apply activation by invoking UseSkillTool with a skill name to disclose the workflow and bundled-resource information to the LLM.
How it connects
Links skill discovery, tool calls, and skill execution through the Agent Skills standard.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Build a Multi-Agent System (from Scratch). The section links point to the parts of the book it draws on.