The third session in the series. A 60-minute working session on Claude skills — what they are, how they work, and how to build your own. Skills are folders of instructions Claude reads when relevant; they turn a repeated request into a one-line trigger.
Audience: anyone who uses Claude.ai or Cowork. No coding required.
Topics covered
- What a skill is — the definition, why it matters, and the two kinds (task-consistency vs. workflow + MCP).
- Inside a skill — folder structure,
SKILL.mdanatomy, the magic description line, progressive disclosure, and optional folders. - Get one. Share yours. — scopes (bundled / personal / shared / org-provisioned) and the two paths to install and to share.
- A real skill, under the hood — a full case study of
/salesforce-churn-check: architecture,SKILL.mdsnippets, the output a CSM gets back, and why references and scripts earn their place. - The workflow, start to ship — build → smoke test (4 checks) → peer review (critique + comparison prompts) → iterate → distribute.
- What’s next — the internal org skills already live, plus a 4-question thinking model for building your own.
The throughline: the body of a good skill is just a polished prompt — which is exactly what Workshop #2 set up.