A skill that makes Claude run a brand discovery interview one question at a time and turn the answers into a one-page positioning document. It reads what you share first, such as reviews, call notes, your site or a client's deck, and asks only what is missing, each question with a proposed answer and its source. Then it describes the audience by the situation they buy in, puts the real alternatives in a grid with a price index, tests candidate points of difference, grades every reason to believe and writes the personality as decisions. It offers two or three territories, you choose one, and it writes the page with its open questions.
Inside are five reference files. An interview guide holds the 15 core questions with weak and strong answers, a client kickoff kit with a 90-minute agenda and a method for sorting reviews and transcripts. A positioning toolkit covers the grid, the category frame, five tests for a difference and the personality scales. A rules file covers US recording consent, paying participants, FTC review and testimonial rules, origin and green claims and comparisons, each sourced and checked on 17 October 2026. There is a full worked example for a fictional cookware brand. Every finding carries a grade from A to D, and Claude never invents a customer, a quote, a price or a competitor claim.
It runs the interview with you, not with your customers. When the audience rests on guesses, it says so, marks those lines as hypotheses and plans the cheapest way to check them, such as tagging 40 reviews or booking five 30-minute customer calls. It flags legal rules but it is not legal advice, and the rules file covers only the US. Like every skill in claude.ai it needs code execution switched on, and that also lets it save a Word file or PDF. It reads competitor sites only when web search is on. You get the most from it by pasting real customer words with the names removed, and by naming the person who signs off before the first stage ends.
Claude skill · version 1.0
Brand discovery
Checking your access.
In the download · 9 files
- SKILL.mdThe method Claude follows: set-up, audience and problem, category and competitors, difference and proof, territories, then the one-pager, one stage and one question at a time.
- README.mdHow to install it in claude.ai or Claude Code, how to start, and the licence.
- references/interview-guide.mdThe 15 core questions with follow-ups and weak and strong answers, a short 8-question version, techniques for getting past the first answer, a client kickoff kit with a 90-minute agenda, customer interviews and how to sort notes and reviews.
- references/positioning-toolkit.mdThree layers of alternative, the competitor grid, choosing the category frame, the price index bands, the five tests for a point of difference, grading reasons to believe, the personality scales and territory cards.
- references/us-research-and-claims.mdUS rules discovery runs into: recording consent, paying participants and the 1099 threshold, FTC review and testimonial rules, origin, green and health claims, comparisons and trademark searches, each sourced and checked on 17 October 2026.
- references/one-pager-format.mdThe one-page format with word budgets, weak and strong versions of every section, the positioning statement, the eight tests, open questions with ways to answer them, the file spec and the hand-off.
Install it and begin
01Install it in Claude CodeTerminal mkdir -p ~/.claude/skillsunzip ~/Downloads/brand-discovery.zip -d ~/.claude/skills/# Start it with /brand-discovery. If this is your first skill, type /reload-skills once.On claude.ai there is nothing to unzip: upload the zip as it is. Every step is on Install and use a Claude skill.
02Your first message to ClaudeText Use the brand discovery skill. The brand is [BRAND], which sells [WHAT IT SELLS] in [COUNTRY]. I am [THE FOUNDER / A DESIGNER WORKING FOR THEM]. The page is for [THE NEXT JOB, FOR EXAMPLE A NEW IDENTITY]. Here is what I have: [PASTE NOTES, REVIEWS OR LINKS]. Read it all, then show me the interview plan.Attach real customer words with the names removed (reviews, support emails or call notes), the brand's product and about pages, and the competitors it loses sales to. If you are working for a client, paste your kickoff notes too.
Try this
Run discovery on a brand you know
Pick a brand you work on, your own or a client's, and run the skill up to the territory choice in one sitting. Paste at least 20 real reviews or customer messages first, with the names removed. The next day, read the one-pager it writes as if you were a designer who has never met the client.
Time: About an hour, plus 15 minutes the next day. You need: 20 to 50 customer reviews, support emails or survey answers, names removed, the three competitors the brand loses sales to, the brand's product and about pages, or the client's brief or deck, the name of the person who signs off.
Look for
- Questions about the last real customer, not the typical one.
- Themes counted, such as 12 of 41 reviews, each with a quote and its label.
- Doing nothing, or the cheap version, as a row in the competitor grid.
- At least one candidate difference dropped because a competitor already claims it.
- Lines marked as hypotheses where the evidence is thin, instead of confident guesses.
- A page you could brief a logo or a shoot from in two minutes.
What to do it with
- ClaudeFree tier, paid above it
A language model that writes, edits, reads long documents and runs tools, including agents that work inside a codebase.
Bad at It is agreeable by default, so it will improve a bad idea instead of telling you it is bad. Ask it to argue against you and the quality of the answer changes.
- Claude CodeSubscription
An agent that reads a whole repository, edits files, runs the tests and explains what it changed.
Bad at It is fast enough to make a large mistake quickly. Work on a branch, read the diff, and never let it push. Watch the starting mode: on Pro, Max and Team plans it opens in auto mode, which Anthropic's own list says may read a .env file and send the key in it to the API that key belongs to. Switch it off.
Why it matters
Ask a chat for brand positioning and it returns a persona with an invented name, three values anyone would claim and a difference every competitor already offers. This skill makes Claude ask before it writes, count what buyers actually say, check each candidate against what rivals already claim and grade every line by its evidence. Weak findings become open questions with an owner, so the identity work starts from facts instead of the founder's favourite adjectives.
- Type
- technique
- Period
- ongoing