Most wasted generations are wasted before the prompt is typed, because the idea behind them is still a feeling rather than a specification. ContentForger has a tool for exactly that stage: the AI Project Brief at /brief. This tutorial shows what it does, what it costs you, and how to use its output so the first real generation lands.
What the brief tool is
You paste a rough idea of between 5 and 2,000 characters. Claude Sonnet, acting under a "senior product architect" instruction, returns a structured brief in a fixed shape:
- App Name: a suggested name.
- Target Users: two or three sentences on who it is for.
- Core Features: five to eight one-sentence bullets.
- Tech Stack: Next.js 16 with TypeScript and Tailwind CSS v4, plus libraries relevant to your idea.
- Estimated Complexity: Simple, Medium or Complex, with one sentence of justification.
- Suggested Prompts: three ready-to-paste prompts for the Generate page.
The output streams in as it is written and renders as a document you can copy. Under it, each of the three suggested prompts appears as a button; clicking one opens /generate with that prompt already filled in.
Who can use it and what it costs
The brief tool is available on the Pro and Agency plans. Free accounts see the page with an upgrade notice rather than the form.
A brief is not a generation. It does not write a row to your history and does not count toward your daily allowance of generations. It does share two other limits with the Generate page: the twenty-requests-an-hour limit per network address, and the site-wide monthly spend ceiling. Those are explained in Plans, Limits and the Agency API Explained.
Step one: write the idea badly on purpose
The input is meant to be rough. Write it the way you would describe the idea to a friend: who it is for, what they do with it, what they pay. Two or three sentences is plenty. Do not try to pre-structure it; that is the tool's job, and giving it a messy input shows you what it adds.
An example of a good bad idea: "A tool for freelance photographers to send clients a private gallery, let them pick favourites, and pay for the ones they want. Free for one gallery, paid for more."
Step two: read the brief as a critique, not a plan
The brief will name the app, describe the user and list features. Read the feature list looking for two things.
Features you did not ask for. The tool fills gaps with reasonable assumptions. Some are right; some reveal that your idea was vaguer than you thought. Strike anything you do not want before generating.
Features it left out. If the one thing that makes your idea different is missing, your input did not say it clearly enough. Rewrite the idea and run the brief again; it costs you nothing from your daily allowance.
In the photographer example, a first brief might list galleries, favourites, payments, an admin view and client accounts with passwords. The client accounts are the assumption to question: a private link may be the whole point, and adding sign-in for clients doubles the scope. Strike it, and the Estimated Complexity line will usually drop a grade on the next run. That is the brief doing its job: not writing your plan, but showing you where your idea was still soft.
The complexity estimate is worth taking seriously. A Complex brief is a signal to build the core first and add the rest through change requests, so each step can be checked before the next.
Step three: use the suggested prompts as drafts
The three suggested prompts are good starting points and rarely perfect finished ones. Before clicking through, edit the one you like against the three habits the Generate page itself recommends: name the stack, list the pages, state the business rules. The brief gives you the stack and most of the pages; the business rules are usually what you have to add. "Free plan: one gallery. Paid: unlimited, $9 a month" is the kind of line that changes what the model builds.
Clicking a suggested prompt carries it to /generate in the address bar. Nothing is generated until you press Generate there, so you can still pick a template and a style first.
Step four: generate narrow, then widen
Use the edited prompt for a first generation that asks for the core loop only: in the gallery example, upload, share link, pick favourites. Leave payment for a follow-up. On Free that is a necessity; on Pro it is still good practice, because a focused first run gives you something to react to within the output budget rather than a truncated everything.
Then follow up inside the same conversation, one concern per message, asking for only the files that change. Tutorial: Iterate on a Generated App With Follow-Up Prompts covers that rhythm in detail.
When to skip the brief
If you already have a written specification, paste that into Generate directly; the brief would only restate it. If you are generating a single component rather than an app, the brief's shape is overkill. The tool earns its place when the idea is real but unstructured, which is most ideas on day one.
What the brief does not do
It does not generate code; that is the Generate page's job. It does not save the brief anywhere; copy it if you want to keep it, because reloading the page clears it. And it does not know your business; it knows what you typed. A brief is a sharp mirror, not an oracle.
For the longer discipline of turning an idea into a prompt that holds up, What Makes a Good Prompt for a Code Generator remains the fuller guide; the brief tool is the fast path through its first half.
