/learn

What it does

Type /as:learn and the course begins — or resumes exactly where you left off. Six modules, each named for the lesson it teaches and each built around one hands-on exercise: your first questions, your first file, your office standards binder, a messy-folder cleanup, building your own skill, and a closing drill in verifying AI output. The tutor guides and reviews; you type the prompts.

Three ideas thread the whole course: your work stays local (files live in your folders on your machine — only what you ask about goes to Claude, like any conversation), Architecture Studio is an open-source harness on Claude Code (fork it, make it your own), and memory is plain markdown files you can open and read.

The practice project

The practice project is a fictional Brooklyn art museum planning a rooftop expansion — the Greenpoint Museum of Art. It ships with the skill as six deliberately messy files, from raw site-visit notes to a half-scanned zoning memo. Among them: a space-program spreadsheet with a duplicate row, mixed units, and a TBD; a voice-memo transcript saved under a photo-style filename; a scanned zoning excerpt that cuts off mid-page. Every module runs on the same project, so by the end you know it well enough to interrogate what the machine says about it.

Install

Claude Desktop:

1. Open Customize → Browse plugins
2. Click + → Add marketplace from GitHub
3. Enter AlpacaLabsLLC/skills-for-architects
4. Install Architecture Studio

Claude Code (terminal):

claude plugin marketplace add AlpacaLabsLLC/skills-for-architects
claude plugin install as@skills-for-architects

Your first session after installing, Architecture Studio confirms the install landed and offers the course on its own.

Usage

/as:learn

That’s the whole interface. First run: pick a practice folder, the museum files are copied in, and Module 1 begins. Any later run: the course reads PROGRESS.md and resumes.

The course

#	Module	What you’ll do
1	How we interact with each other	Ask in plain English; it reads your files
2	Nothing without your “yes”	Your first file · approvals · your data
3	Let’s set some guidelines first	Your standards binder (CLAUDE.md)
4	Plan first, build second	Messy files → order, on a plan you edited
5	Creating your own skills	Build your own /site-report command
6	Get started	Verify like a pro, then your real project · 30–60 min

Module 6 is the point of the whole course: the tutor summarizes the museum’s zoning excerpt faithfully, and you interrogate it — pick claims and ask “show me where in the document it says that”, tracing each answer to exact lines. Then you ask about a topic the excerpt never covers and hear the only honest answer: the document doesn’t say. Silence and “no requirement” are different things, and that question works on any AI output, forever. The module closes with a short professional checklist — your firm’s policy on AI and client material, something low-stakes before anything that matters, work on a copy — and one real task on a project of your own.

Data privacy is taught where it belongs: module 2, the module about consent, ends with what left your machine (nothing but what you asked about — no ALPA servers, no cloud copies) and the forward rule: know your firm’s data-governance policy before pointing the tool at client material.

An advanced track — bigger jobs, plan mode, subagents, running the office skill library — is planned as a follow-on course.

Design notes

• Nothing is written without your yes. The whole course runs on the permission model it’s teaching; the sandbox is fictional by design, so there is genuinely nothing to break.
• Progress is a file. PROGRESS.md in your practice folder tracks completed modules — which is itself the lesson in how memory works here.
• Stop or quit anytime, zero friction. Every module ends with the same return ritual, and asking to quit — even mid-module — stops the course immediately, no persuasion. /as:learn picks up where you stopped, across sessions and days.