The brief-builder skill: turn a vague ask into one you can paste
TL;DR
The brief-builder skill takes a vague ask and turns it into a 5-Part Brief you paste straight into Claude: context, goal, constraints, format, and done-when. It saves the brief to ~/briefs/ and hands you a one-sentence compressed version for small asks. Run it before any real build.
Ask Claude for a website and you get a website. It just won't be yours. That happens because Claude can't see your life: your customer, your price, the thing your audience actually complains about.
The brief-builder skill closes that gap before you build. You describe what you want however messy it comes out, and it turns your ramble into a 5-Part Brief: context, goal, constraints, format, and done-when. You paste the whole block into a fresh session and run it. Same frame works for a landing page, an email, or a spreadsheet.
The part beginners feel immediately is done-when. It's the line that says how you'll know the thing is finished, checkable by looking at something rather than by feeling satisfied. That one line is what stops the endless tweaking loop. The skill is free and it lives in our public freeskills repo.
How it finds the deliverable, fills the five parts, and rewrites a done-when you can check by looking.
Install the brief-builder skill
Clone the repo, then copy this folder into your Claude Code skills directory.
bashgit clone https://github.com/duncan-buildroom/freeskills.git
mkdir -p ~/.claude/skills
cp -r freeskills/brief-builder ~/.claude/skills/Restart Claude Code or start a fresh session and it's ready. No API keys, nothing to configure.
Run it the first time
Type /brief-builder and describe what you want in your own words. It reads your ask back as one sentence naming the thing that exists at the end, and waits for a yes before writing anything. If you named three things, it stops and makes you pick the one you'd do first, then parks the rest as next briefs.
Then it asks only for the details that change the answer, and writes the brief to ~/briefs/<slug>.md. Inside that file: your original ask word for word, the five labeled parts in a copy-ready block, the compressed one-sentence version, and anything you parked. Honest expectation: if you give it two sentences and no detail, it won't run a twenty-question interview. It fills in sensible defaults, marks them as assumptions at the bottom, and lets you correct one.
Turn tonight's build into a brief
Say it exactly as messily as you'd say it out loud. That's what it's built for.
/brief-builder
I want [what you want, in your own words, however vague].
It's for [who, one describable person, not everyone].
What already exists: [files, a site, a doc, or nothing].
Things I know that Claude can't guess: [your price, your offer, what your customer complains about].Pin down the constraints when you can't name any
Constraints are what make the output specific, and most people freeze when asked for them. Ask it to propose and let yourself react instead. Rejecting is easier than inventing.
I don't know what my constraints are.
Give me a few you'd guess for [the thing you're building] and I'll tell you which ones are right.
Things I already know I hate: [a look, a tone, a word, a color].Run the brief, then change one part
If the result comes back wrong, that isn't a failure, it's the loop. Change one part of the brief and run it again.
I pasted the brief in ~/briefs/[your-slug].md and here's what came back: [describe what you got, or paste it].
The part that's wrong is [what's off].
Change ONE part of the brief so the next run fixes it, and tell me which part you changed and why.The 60-second version
Here's what it does and why it fixes your results, in a version you can say in a minute.
- 1Hook: there's a brand new free Claude skill that takes you from asking for a website and getting a generic one to a brief that gets you the right thing on the first try.
- 2Mechanism: you ramble at it about the thing you want, and it pulls the actual deliverable out of the mess. Then it fills in the five parts you'd never have written yourself, the context that's true right now, the one goal, the constraints, the shape of the answer, and how you'll know it's finished. It hands back a full brief plus a short version for everyday asks.
- 3Transformation: you stop blaming the model for thin answers. You get a saved brief you can rerun and edit, and one shorter version you'll use forty times a week.
- 4Close: the whole repo is free, 55 skills in it. Knowing what's worth briefing in the first place is the part we work on inside Claude Code Club, $9 a month. Link below.
Tonight: run it on the thing you've been putting off building, paste the brief into a fresh session, and let it go.
Common questions
Do I need to know how to code?
No. The skill never puts code in a brief. You describe the outcome and Claude picks how to do it. That's the whole arrangement.
Is it really free?
Yes. It's public in the freeskills repo, needs no API keys, and saves plain markdown files to your own machine.
What if I want three things at once?
It'll stop you. One brief, one deliverable. It helps you pick the one you'd do first, then parks the rest as next briefs at the bottom of the file.
What if what I want is huge, like a whole app?
It writes the brief for the first shippable slice only and lists the rest as next briefs. You get something real out of the first run instead of a plan you never start.
I already ran a prompt and got something bad. Should I use this?
Use prompt-fixer from the same repo instead. It diagnoses why the prompt failed before anything gets rewritten.
Keep going
Want the whole ask-build-ship path?
Get 650+ plug-and-play skills, MCPs & prompts, plus 7,000+ members - $9/mo, cancel anytime.
Join the Club