The Manager's AI Toolkit
↓ Download

Workshop Starter Project

A ready-to-use project folder for an AI workshop. Copy it, spend ten minutes filling in four files, and your team starts the workshop with working skills and commands instead of a blank window.

Hand this out at the kickoff meeting, one week before the workshop.

⚠️ This one needs Claude

CLAUDE.md, .claude/skills/, and .claude/commands/ are Claude features. This folder works in the Claude VS Code extension, the Claude desktop app, claude.ai, or the terminal — and nowhere else. There’s no version of this that Copilot or Gemini reads.

If your company approved something other than Claude — which is most companies, since Copilot and Gemini arrive bundled with Microsoft 365 and Google Workspace — use prompt-recipes.md instead. It’s the same seven jobs (meeting notes, recurring reports, reply drafts, document briefs, plain-English explanations, critique, checklists) as copy-paste prompts that work in any assistant. You lose the automatic context loading and gain the ability to actually run it. Hand that file out at the kickoff exactly the way this README describes, and the rest of the workshop is identical.

Not sure what you have? Check what’s already paid for before you plan around this folder.

What’s in here

workshop-starter/
├── README.md              ← you are here
├── CLAUDE.md              ← project instructions; the important file
├── context/
│   ├── team.md            ← who we are and what we work on
│   ├── voice.md           ← how we write
│   ├── data-rules.md      ← what must never be pasted in
│   └── glossary.md        ← our jargon, systems, and acronyms
└── .claude/
    ├── skills/            ← multi-step workflows, invoked by describing the task
    │   ├── meeting-notes/SKILL.md
    │   ├── weekly-report/SKILL.md
    │   ├── reply-draft/SKILL.md
    │   └── doc-brief/SKILL.md
    └── commands/          ← one-shot slash commands
        ├── plain-english.md
        ├── critique.md
        └── checklist.md

Skills vs. commands. A skill is a workflow the assistant reaches for on its own when the task matches its description (“turn these notes into actions”). A command is something you invoke deliberately by name (/critique). Skills are for repeatable jobs with several steps; commands are for quick one-shot operations on whatever you’re looking at. Under the hood they’ve converged — a file in either folder gives you a /name you can type — so if you’re unsure which to write, write a skill.

Setup (10 minutes, do it in the kickoff meeting)

  1. Copy this folder somewhere your team can each have their own copy — a shared drive folder, a repo, wherever your team keeps things.
  2. Fill in context/data-rules.md first. It should match your company’s actual policy. Don’t guess — this is the file that makes everything else safe to use.
  3. Fill in context/team.md — five minutes, as a group, on the projector. It’s the single highest-value file in here.
  4. Fill in context/glossary.md with your ten worst acronyms.
  5. Skim CLAUDE.md together and delete anything that isn’t true for your team.
  6. Test it. Paste a real (non-sensitive) set of meeting notes and ask for action items. If the output uses your team’s names and terminology, the context files are wired up correctly.

context/voice.md can wait — fill it in when someone’s first draft comes back sounding wrong.

How to use it during the workshop

Pick the row that matches your team. Everyone can use this folder; what differs is how much of it is picked up automatically.

You’re usingSetupSkills and commands
VS Code (Claude Code extension)Open this folder as the workspaceAutomatic — nothing to upload
Claude desktop app or claude.aiMake a Project, or upload skills onceUpload each skill as a ZIP, or paste it
Terminalcd into this folder, run claudeAutomatic — nothing to upload

Full instructions for each below.


VS Code

The best option if your team already lives in an editor, and the one where this folder works with zero setup.

  1. Install the extension. In VS Code press Cmd+Shift+X (Mac) or Ctrl+Shift+X (Windows), search for Claude Code, click Install. VS Code 1.94 or later.
  2. Sign in. Open the panel and click Sign in. Any paid Claude plan works — Pro, Max, Team, or Enterprise. No API key needed.
  3. Open this folder as the workspace. File → Open Folder → select your copy of workshop-starter. This is the step that matters. Everything below is discovered relative to the folder you have open, so if you open a parent folder or no folder at all, none of it loads.
  4. Open the Claude panel. Click the spark icon in the top-right of the editor (it appears when a file is open), or click ✱ Claude Code in the status bar at the bottom-right, or press Cmd+Shift+P / Ctrl+Shift+P and type “Claude Code”.

What gets picked up automatically, with no upload step:

  • CLAUDE.md — including the @context/... files it imports
  • .claude/skills/ — all four skills
  • .claude/commands/ — all three slash commands

Using them:

  • Commands: type / in the prompt box to see the menu, then pick one — /critique, /plain-english, /checklist.
  • Skills: just describe what you want (“turn these notes into action items”) and Claude reaches for the matching skill. You can also invoke one directly by name: /meeting-notes.
  • Files: type @ and part of a filename to point Claude at something specific. Selected text in the editor is visible to Claude automatically.

Editing a skill takes effect immediately — save SKILL.md and the next message uses the new version, no restart. That makes the edit-and-retry loop fast, which is the loop the workshop is really teaching. (If you create a brand-new top-level folder, restart the extension so it starts watching it.)

One caveat: the VS Code panel offers a subset of what the terminal version does. Type / to see what’s available.


Claude desktop app or claude.ai

No install, no editor. Two ways to use this folder, and they’re worth understanding separately.

Option A — a Project (simplest, start here).

  1. Create a new Project.
  2. Paste the contents of CLAUDE.md into the project’s custom instructions. Delete the “Referencing other files” section — it describes a mechanism that only applies to the editor and terminal versions.
  3. Upload the four context/*.md files as project knowledge.
  4. When you want a particular workflow, paste the body of that SKILL.md into the chat as your instructions. The content is what does the work; the folder structure is just packaging.

Every chat you start inside that Project now knows your team, your voice, and your data rules.

Option B — upload the skills properly (do this once, keep it forever).

The apps support custom skills directly, so you don’t have to keep pasting:

  1. Enable code execution in settings — required for skills on every plan.
  2. ZIP up one skill folder (for example, zip the meeting-notes folder so the ZIP contains its SKILL.md).
  3. Go to Customize → Skills, click +, choose Create skill, then Upload a skill, and select the ZIP.
  4. Repeat for each skill you want.

Claude then applies them automatically when a conversation matches — you don’t have to invoke anything, though you can nudge it (“use my meeting-notes skill”). Available on Free, Pro, Max, Team, and Enterprise plans. Uploaded skills are private to your account; Team and Enterprise admins can share them organization-wide.

What doesn’t transfer: the .claude/commands/ files are a Claude Code feature. In the apps, paste the body of the command as your message instead — it’s written to work that way.

Which option? Use a Project for the workshop itself; it takes two minutes and everyone can do it at once. Upload skills afterward for the ones people actually keep using.


Terminal

If your team is comfortable at a command line: cd into your copy of this folder and run claude. Everything — CLAUDE.md, the context files, the skills, the commands — loads from the working directory, same as the VS Code extension, and the full set of commands is available.

Adapting it

These skills are deliberately generic. The point of the workshop is that people replace them with their own. Encourage participants to:

  • Edit a skill’s steps when the output isn’t what they wanted. That edit is the skill-building.
  • Add a new skill for their workshop project — copy the closest existing one as a starting shape.
  • Add to context/ whenever they find themselves explaining the same background twice.

At the end of the workshop, merge everyone’s best additions back into one shared copy. That merged folder is your team’s prompt library.