Project-Manager-Package
A Cowork / Claude Code plugin marketplace that ships two things:
/project-manager — a tool-agnostic AI project-management skill (set up, organize, and run software projects from intent to shipped artifact)./anchor-coach — the Anchor curriculum, an interactive first-project course that teaches an absolute beginner to build, test, and ship a working web app by directing an AI agent through eight milestones.
The two are designed to work together: the Anchor curriculum quietly uses /project-manager under the hood, and the final lesson reveals this and hands /project-manager to the student to use directly on a project of their own choosing.
Getting started — for students
Three steps. About 5 minutes from zero to your first lesson.
1. Install Cowork
Go to claude.com/cowork, download for your OS, and sign in with your Anthropic account.
An Anthropic subscription is required. Free-tier limits may apply; check the current Cowork docs.
2. Add this marketplace and install the curriculum
Open Cowork. Paste these three commands into the chat, one at a time, and press Enter after each. Wait for the confirmation between each one.
/plugin marketplace add johncliechty/Project-Manager-Package
/plugin install project-manager@project-manager-package
/plugin install anchor-coach@project-manager-package
3. Start your project
Make an empty folder anywhere on your computer (right-click your Desktop → New Folder → name it anchor or whatever you want). In Cowork, click Open Folder and pick that folder. Then type:
Let's start the Anchor curriculum.
Press Enter. The Anchor coach takes it from there. The first lesson is about 10 minutes. The whole curriculum is about 2 hours of session time, spread across whatever calendar feels right. You'll finish with a working web app shipped to your own public GitHub, a retrospective about your work, and a second project already bootstrapped.
What you'll build
Anchor is a single-file local task tracker:
- One
index.html, vanilla JavaScript, localStorage for persistence — no toolchain, no server, no install.
- Six features: add a task, mark complete, delete, group tasks under projects, persist across page reloads, polish (system font, comfortable whitespace, accent color, anchor SVG).
- Tests written by the agent that prove every feature works.
- Public GitHub repo with a README and screenshot at the end.
The course teaches the loop you'll use for the next twenty years of building things with an AI agent. Anchor is the seed; the loop is the plant.
The 8-milestone spine
| # | Milestone | What you produce | Time |
|---|
| M0 | Setup | Cowork running, GitHub auth, project folder | ~10 min |
| M1 | Brainstorm | brief.md — who it's for, scenarios, NOT-list | ~10 min |
| M2 | Describe | picture.md — mockup, 5 data nouns, 4 interactions | ~15 min |
| M3 | Scaffold | index.html empty shell — agent shows 3 mockups, you pick | ~30–40 min |
| M4 | Build | Working index.html with all 6 features + tests | ~10 min |
| M5 | Test | Use it for real; find 2–3 things; fix them | ~10 min |
| M6 | Ship | Public GitHub repo + README + screenshot + text the link | ~10 min |
| M7 | Launch | Retrospective + meta-reveal + your next project bootstrapped | ~25 min |
Total: ~2 hours of session time. The course assumes you spread it across whatever calendar feels right.
For instructors and developers
The curriculum is designed to be remixed for other first-project shapes (Business Statistics tool, data pipeline, content tool, etc.). The seven principles, the milestone spine, the rubric shape, the git contract, and the coach contract are reusable; what changes per course is the feature checklist and the technical scope. See courses/anchor-coach/lessons/ for the lesson format reference and PLAN.md (in the Teaching repo) for the design rationale.
Repo structure
affaan-m
obra