Build One Work Card

What this is

Work Card Build is the implementation loop. The coding agent reads the saved plan, completes only the current Work Card, verifies it, updates the project status, and stops.

Why it matters

Small cards keep the build understandable and recoverable. If a chat ends or you switch models, build-status.md tells the next agent exactly where to continue.

What to do

After planning is complete, you may switch from a strong planning model to a cheaper or faster implementation model.

Before every Work Card, check that design.md exists and is specific. If it is missing or vague, stop and return to Design Direction.

Paste this implementation prompt:

Read build-status.md, build-blueprint.md, design.md, and the current work card listed in build-status.md.

Implement only the current work card.

Rules:
- follow the approved design.md
- do not copy external brand identity, logos, testimonials, stats, or images
- do not start the next work card

After finishing:
1. run the verification steps,
2. summarize what changed,
3. update build-status.md,
4. stop.

For every card, check this loop:

read status -> read blueprint -> read current card -> implement -> verify -> update status -> stop

Do not ask the agent to continue automatically. Read the result and verification first. When you are satisfied, start the next card from the updated build-status.md.

If a chat or credit limit interrupts the build, open a new session and use the same implementation prompt. The files, not the old chat, provide the continuation state.

Common mistakes

• Implementing several Work Cards in one uncontrolled pass.
• Forgetting to run the card's verification steps.
• Updating code but not build-status.md.
• Trusting a completion claim without checking the visible result.
• Asking a new model to remember the old chat instead of reading the files.

Next step

After the build and Review Mirror pass, prepare your Proof and Sharing.