C3. Skills and agents

Outcome: you can package a lab chore as a Claude Skill someone else can use, and explain where subagents and MCP fit.

Skills: what a Skill is (a folder with a SKILL.md: a name, a description of when to use it, and steps), writing a description that makes Claude pick the Skill at the right time, checkable steps, adding a template or script, testing in fresh chats, sharing; subagents for splitting big tasks; MCP as the way Claude connects to other tools and data.

Courses: Introduction to agent skills, Introduction to subagents, Introduction to Model Context Protocol (the first lessons are enough unless you write Python).

Labs

1. Pick a chore you repeat in the lab, for example the snapshot checklist in LAB-ENVIRONMENT.md or the write-up for a sprint's "prove it".
2. Write the Skill: a SKILL.md with a name, a description and numbered steps, plus any template file it needs. Upload it in Claude (Customize, then Skills, with code execution on), or add it to your C2 repo for Claude Code.
3. Test it: in a fresh chat, ask for the chore without naming the Skill. Does Claude use it? Fix the description until it does, and check that it stays out of unrelated requests.
4. Hand it over: a partner installs your Skill and uses it from its instructions alone. Fix whatever confused them.
5. Subagents (Claude Code; skip on the free plan): ask Claude Code to review three of your C2 scripts in parallel with subagents, then compare their findings with your own review.
6. MCP: read how Claude connects to a tool you already use (GitHub or a file system, for example). List the data it could reach, and decide what you would and wouldn't connect at work.

Prove it: the Skill folder in your repo, a test log (requests that should and shouldn't use it) and your partner's notes from lab 4.

Certs: none.