How do Agents and Skills interact in the Claude Code Agent short course?

Short Course Q&A Agent Skills with Anthropic
1

image
image924×1662 89.7 KB

Hi everyone,

I’m going through the Agent Skills section in the Anthropic short course (last lecture: Skills with Claude Code Agent), and I’m a bit confused about how the agents and skills are connected internally.

In the example:

  • First, we created an orchestrator agent

  • Then we created 3 sub-agents

  • After that, we defined skills

However, I noticed that:

  • The skills definitions don’t seem to explicitly reference any sub-agents or tools via a path.

  • The sub-agents also don’t appear to have any direct reference to the skills.md file.

So I’m trying to understand:

  1. How exactly are the agents and skills interacting?

  2. What is the sequence of input → processing → output?

  3. How does the output of a skill get consumed by an agent?

  4. Is this linkage happening implicitly through the framework runtime?

If someone could explain the control flow (maybe with a simple example of how input flows from orchestrator → sub-agent → skill → back), that would really help.

1 Like