When should you use it?
- You have a finished Figma design
- The client needs to edit content themselves
- You will review and finish the build
Not for: Sites without a design yet; Complex custom plugins
What do you need, and what do you get?
You need
- Figma design the agent can read through MCP
- WordPress hosting or local setup
- List of pages and editable fields
You get
- Themed WordPress site
- Editable content blocks
- List of gaps to finish
How does it work, step by step?
- Prepare the designName frames and components clearly in Figma.
- ConnectGive the agent read access to the file through MCP.
- Brief itList pages, which content must be editable and what to skip.
- Build overnightLet it build on a staging site, not the live one.
- ReviewCheck each page against the design and on a phone, then finish the gaps.
The prompt
Our own prompt, inspired by bnistordev (@bnistordev) and written to work in Custom agent. Values in lime are placeholders.
Fill in the blanks (optional; the copied prompt uses your values)
Read the design in {figma_file} and build it as WordPress pages on {staging_url}: {pages}. Make {editable_fields} editable by a non-technical user. Do not touch production. When done, list each page, anything that differs from the design, and what is unfinished.
Inspired by bnistordev (@bnistordev): Probably the first time an agent has given me output this good. The company I work with needed a new marketing site,.... The prompt above is written by playbooks.bot in our own words; nothing from the original is quoted.
What can go wrong?
- Overnight output is not done; plan for finishing work.
- Check accessibility and page speed yourself.
- Keep the agent off the production site until you approve.
Our take
The interesting part of this post is the connection, not the overnight timing. Through MCP the agent reads the actual Figma frames, names, spacing and components rather than guessing from a screenshot, which is why the output resembled real client work. To get similar results, tidy the design file first: name layers sensibly, mark which sections repeat, and include mobile frames, since agents copy whatever structure they are given. Brief the agent on the editing model as well as the look, for example which blocks the marketing team must be able to change without a developer, because a pixel-perfect theme that nobody can edit fails the actual request. Keep the build on a staging site, review it on real phones, run an accessibility checker, and test the editor as the person who will update content. The poster's pricing remark is a useful frame for freelancers: if an agent drafts the build, the value you sell shifts toward design judgment, content structure, performance and the final round of fixes, which is exactly what remains unfinished here.
Where the time goes back: A first full build of a marketing site appears overnight, ready for your finishing pass.
FAQ
Was it finished?
How good was it?
Source and credit
Inspired by bnistordev (@bnistordev). We describe the idea in our own words, link to the original, and wrote the prompt ourselves; nothing is quoted.
- Inspired by
- Probably the first time an agent has given me output this good. The company I work with needed a new marketing site,...
- Creator
- bnistordev (@bnistordev)
- Published on
- X
- Original date
- Sep 29, 2026
- License
- Not openly licensed (the poster keeps all rights), so we quote nothing and link to the original
- Originally for
- Custom agent
- AI dry run
- Oct 23, 2026: passed. Dry-run tested by our AI test agent on sample input. PASS: Gates held on desk check; no live sends, bookings or purchases.
AI dry run on sample data; not run with live connected accounts. The playbooks.bot test agent filled the prompt with synthetic inputs, produced the output and checked it against the page's stated outputs and guardrails. - Verified
- Oct 23, 2026: source link resolves to the original, the credited creator published it, our excerpt or prompt matches it, and the license allows our use.
Are you the creator? Claim or correct this page.