In my last Indie Hackers post, I talked about building Workflow Place -- an Obsidian-style directory for form automation recipes. The design decision I was most uncertain about was deleting the apply modal and replacing it with copyable prompt templates.
This post is about the first recipe I published there, and the product thinking behind it. The recipe turns a client hearing sheet into a Figma wireframe. But the real story is about packaging tacit professional knowledge as a product feature.
Before building FORMLOVA, I was a freelance designer and PM. Over 100 website projects on CrowdWorks, Japan's largest freelance platform. 4.9 rating. Selected for the Top Class Directory and Pro CrowdWorker status.
Every project started with a client hearing. And every hearing had the same problem: the questions I needed to ask were predictable, but the process of asking them was not. Zoom calls, Slack threads, email follow-ups. Information scattered across three channels, consolidated manually into meeting notes, before any design work could begin.
I had a private hearing sheet -- a document I refined across 100+ projects -- that covered business context, target audience, content, design direction, and assets. 28 questions, ordered to match how a designer thinks. It was not a form. It was a Google Doc I walked through verbally.
When I started building FORMLOVA (a chat-first form service with 127 MCP tools), I turned that document into a structured multi-page form. Five steps, 28 questions, with dropdowns where free text would produce vague answers.
Then I connected FORMLOVA and Figma MCP servers and wrote a prompt that reads the hearing responses and builds a wireframe automatically. 28 answers in, wireframe out. About 7 minutes total.
And then I had an interesting product question: is this a blog post, or is this a feature?
The hearing-to-wireframe workflow has three components:
I could have just written about this. "Here is a cool thing you can do with MCP." That would have been fine.
But I realized it fit exactly into the Workflow Place model I had already built. A reusable automation recipe, published as a prompt template, that anyone can copy and run. The infrastructure was already there.
So both recipes live on Workflow Place now. Copy the prompt, paste it into your MCP client, run it. The hearing form appears. Send the URL to a client. When they respond, run the second prompt. A wireframe appears in Figma.
The product decision was: treat domain expertise as content that lives inside the product, not outside it.
Here is what I think is the underrated insight.
The 28 questions in that hearing sheet are not generic. They are the output of 100+ projects' worth of learning about what makes a good first draft. Three questions matter more than the rest:
These are not obvious questions. A junior designer would not know to ask "what do you want to avoid" before "what do you want." A developer building their first landing page would not know that the first-view layout choice determines 60% of the wireframe structure.
This is tacit knowledge. It lives in experienced practitioners' heads and gets transmitted informally through mentorship, if it gets transmitted at all.
Workflow Place turns tacit knowledge into copyable prompts. A designer with 100 projects of experience saves their workflow. A designer with 5 projects copies it and gets the benefit of those 95 additional projects' worth of question design.
Industry, mood, font direction, first-view layout, and primary conversion action are all dropdown selectors. When I used free text for "industry," clients wrote "IT-related," "technology," or "SaaS" for the same category. Dropdowns produce consistent, machine-readable data.
Machine-readable matters because Recipe 2 maps dropdown values directly to wireframe parameters. "Modern & tech-forward" maps to 12px border radius, generous spacing, and dark sections. "Warm & friendly" maps to 16px border radius and soft backgrounds. The LLM does not need to interpret -- the mapping is deterministic.
Business context before design preferences. Target audience before content. Content before visual direction. Assets last.
Asking about colors before understanding the business produces superficial answers. The form enforces the right thinking sequence, which is something a verbal hearing cannot guarantee.
One page with 28 questions would cause immediate abandonment. Five pages with 5-7 questions each feel manageable. This is basic form design, but it matters enormously for completion rates.
I tested with GreenLeaf Analytics -- a fictional AI-powered SaaS for e-commerce cart recovery. The hearing responses included a headline, five business metrics, green as the main color, split first-view layout, and 10 sections selected.
The generated Figma wireframe: GreenLeaf Analytics - LP Wireframe
Of the 28 hearing items, 14 mapped directly to wireframe elements. The headline became the hero heading. The CTA text appeared in three locations. The color showed up in buttons and accents. The first-view selection determined the hero layout direction.
The remaining 14 items were used indirectly -- business description generated FAQ questions, target audience influenced section copy, competitive advantages shaped the solution narrative.
Total time: about 7 minutes (2 min form generation + 1 min test response + 3 min wireframe generation). The traditional flow -- 1-hour hearing, 30-minute notes, 3-4 hours wireframe -- takes at least half a day.
This hearing sheet recipe is one example of a pattern I think will become common in the MCP era: domain expertise packaged as prompt templates.
Think about it. Every profession has workflows where:
Sales discovery calls that should produce qualified lead profiles. User research interviews that should produce personas. Project kickoff meetings that should produce scope documents.
In each case, the expert's value is not in executing the output -- it is in knowing the right questions and the right order. That knowledge can be encoded in a structured form + a generation prompt.
Workflow Place is the distribution layer for this pattern. The recipe is the packaging. The prompt is the interface.
Since the last Indie Hackers post:
For those of you building in the MCP space or in professional services:
What tacit knowledge in your domain could be packaged as a structured form + generation prompt? What are the questions that experienced practitioners know to ask but never write down?
I have a feeling this is a larger opportunity than it appears. The MCP ecosystem provides the orchestration. Forms provide the structured intake. Prompt templates provide the packaging. The missing piece is the domain expertise -- and that is exactly what individual practitioners and small teams have in abundance.
The hearing sheet was mine. What is yours?