KDBM Lite · 09 / 10

Build in Your Workspace

Let the KDBM Lite Coach infer the build shape, save a checkable spec, and implement one Work Card at a time.

Goal

Turn your idea into durable project files and build one Work Card at a time.

Do now

Confirm the build shape, approve the files, implement only the current Work Card, then run Review Mirror.

Output

Project files, one working build slice, browser evidence, and updated build status.

Stop when

The current Work Card works in the browser and the Review Mirror passes.

Build in Your Workspace

What this is

This is the KDBM Lite build loop. The coach turns your idea into durable project files, then you build one small Work Card at a time.

Copy

Idea
  ↓ infer + confirm build shape
project-brief.md → architecture.md → design.md
  ↓
build-blueprint.md → build-status.md → work-cards/*.md
  ↓
Build one card → Review Mirror → update status → next card

Treat this as a chain, not a checklist. Each file gives the next agent less room to guess.

These files are the only progress source. The Academy does not track your build.

Why it matters

The coach can adapt to what you are making without forcing you to choose a course label. It states its interpretation, asks you to confirm once, then applies the right checks.

Inferred build shape	Required guardrails
Content-led website	Clear sections, CTA, factual content, mobile layout
Browser-local tool	Data shape, add/update/delete/save, refresh persistence
Ambiguous idea	One clarifying question before planning
Backend, auth, payments, database, or live API	Scope down unless the trainer explicitly approves it

What to do

Do this now

Steps 1 to 5 plan the build. Steps 6 to 8 are the one-Work-Card loop you repeat.

Describe the user, problem, and smallest useful result.
Confirm or correct the coach’s build-shape interpretation.
Answer one question at a time while the coach saves each planning file.
Approve design.md only when it contains specific visual, mobile, and do-not-copy rules.
Review build-blueprint.md, build-status.md, and the first Work Card.
Tell the agent to implement only the current Work Card.
Run the Review Mirror: compare the result with the brief, design, acceptance checks, and real browser behaviour.
Record the evidence and next action in build-status.md before continuing.

If the chat resets, tell the next agent to read START_HERE.md and build-status.md, then resume the current Work Card.

Check yourself

I am building one Work Card, not several.
I used the main action myself in a browser.
The Review Mirror checked the result against the brief and design.
build-status.md records what passed and what is next.

You are ready when

The current Work Card works in the browser and the Review Mirror passes. Browser evidence makes a card done, not the agent saying it finished.

Common mistakes

Accepting the wrong build shape without correcting it.
Using words like “modern” or “clean” instead of specific design rules.
Building several Work Cards before checking the first one.
Marking a card complete without browser evidence.
Adding login, payments, or a database during a beginner session.

Next step

When the main action works and the Review Mirror passes, Finish & Share.