Claude Mastery Path — Five-Level Practical Curriculum
For public workshops and open courses. Built on our team’s own 16-module
claude-courseskeleton, deepened and restructured into five levels. Design principles: 1) build on existing, field-tested material; 2) every teaching unit ships with a reproducible Demo Pack (any instructor can run it live on stage); 3) every claim links back to real material insources/. Created: 2026-06-26 | Status: in progress (L1 template first)
1. The Five-Level Map
| Level | Name | Focus | Exit competency | Maps to claude-course modules |
|---|---|---|---|---|
| L1 | Foundations | Chat fundamentals · Prompting framework · Artifacts · Safety awareness | Complete 3 types of everyday tasks using the prompting framework | Meet Claude, Your First Conversation, Prompting Essentials |
| L2 | Personal Workflow | Files & multimodal · Artifacts · Design · Projects · Connected tools | Set up your own workspace + a personal SOP | Uploading Files, Artifacts, Design Work, Projects, Connecting Tools |
| L3 | Agentic Work with Cowork | Deliverable mindset · Computer & file operations · Cross-source research · Permission safety | Delegate a task and receive a complete deliverable back | Claude Cowork |
| L4 | Engineering Automation with Claude Code | Claude Code · Plan Mode / Subagents / Hooks · MCP · Skills · Automation | Direct work in natural language + automate one full workflow | Code Basics, Core Workflows, Skills, Memory Customization (CLAUDE.md), Automation |
| L5 | Advanced Integration & Org Adoption | Deeper subagents & MCP · Claude in Slack (tag @Claude) · Governance · ROI · Adoption | Design a team AI workflow + an adoption roadmap | (extends the claude-course “power-user tips” track) |
Cross-cutting threads (span all levels): safety (see 07_限制安全與失敗案例庫, the Limitations, Safety & Failure Case Library), six skill modules (presentations / integrations / extending / web data extraction / professional documents / UI & settings), the glossary quick reference, and instructor tips.
Prep resources: Teaching Radar Highlights/ — consensus points, demos, and quotable lines distilled from 45 expert videos, each tagged with “which level it feeds into.” Use it when prepping to pick teaching approaches or find a great one-liner.
The progression arc: personal use → personal mastery → delegating to agents → engineering automation → organizational scale.
2. Fast Tracks by Role
| Role | Recommended path | Key labs |
|---|---|---|
| Office & Admin | L1 → L2 → (L3) | Document summaries, slide decks, meeting action items, cross-file consolidation |
| Sales | L1 → L2 → L3 | Prospect research, proposals, follow-up emails, client meeting prep |
| Nonprofit Advocacy | L1 → L2 → L3 → (L4) | Policy research, press releases, data storytelling, public-hearing prep |
| Founders | L1 → L2 → L4 → L5 | Market research, MVPs, marketing content, automation, fundraising |
3. The Demo Pack Standard (required for every teaching unit)
The goal: any instructor can run the demo live, unassisted. Every demo follows the same fixed fields:
| Field | What it covers |
|---|---|
| Goal | One sentence: what learners should see or learn from this demo |
| Difficulty / ⏱ Time | Beginner / Intermediate / Advanced; 5-, 15-, or 45-minute versions |
| Prerequisites | Account plan, tools, and where it runs (Chat / Cowork / Claude Code) |
| Test data | Always dummy data, with instructions for generating it; never touch real confidential material |
| ⌨ Copy-paste prompt | Localized, real-world scenarios — ready to paste and run |
| Expected output | What the result should look like when it works |
| Failure signals & fallback | Warning signs the demo went sideways + how to recover live (including when the network dies) |
| Wrap-up line | One sentence to land the takeaway after the demo |
| Source material | Which file or video transcript in sources/ to go deeper |
Which demos can be pre-generated? (this curriculum’s rule of thumb)
- Pre-generatable: sample HTML Artifacts, example prompts with expected outputs, sample documents → stored in
LX/demo包/so instructors can show them directly or use them as a reference. - Requires live operation: anything that must be clicked through in the real claude.ai / Cowork / Claude Code interface, or needs a real account or paid feature → written up as a step-by-step runbook and logged in
待處理_需你操作.md(the “pending: requires live operation” checklist) with full how-to steps.
4. Lesson Structure per Level
Measurable learning objectives → core modules → multiple Demo Packs → four-track hands-on labs → anti-patterns & common mistakes → capstone project → competency rubric → bridge to the next level.
5. How to Use This Curriculum
- Instructor prep: read the level’s
LX_教案.md(lesson plan) → dry-run the pre-generatable demos indemo包/→ work through the live-operation demos per待處理_需你操作.md. - Teaching: pair with
05_教案/工作坊簡報.pptx(workshop deck) and theDemo_Runbook. - Verification discipline: re-verify all star ratings, prices, plans, and third-party figures before every class (see the source ledger at
index/來源台帳.md).
6. Progress
See 00_進度.md (the progress tracker).