Four personal-brand prompts — symbol, slogan, unexpected idea, focus — only work if you run them one after another without losing the context between them. Bundled into a single skill they turn into a process Claude runs itself: it asks the questions, walks the steps in the right order and checks the result at the end. Below is how to build it, step by step.
What a skill adds to ordinary prompts
A skill is a SKILL.md file with a name field (how Claude names and invokes it), a
description (which request it applies to) and a body with the instructions themselves. The
difference from a one-off prompt isn't technical but practical: you don't have to explain the format,
the order and the criteria every time — the skill holds them and won't let you skip a step.
Step 1 — describe when the skill should fire
The short description at the top of the file is what tells Claude to apply this particular skill instead of answering in general terms.
---
name: lichnyj-brend
description: Builds a personal brand step by step — symbol, slogan,
an unexpected niche idea, positioning focus
---
Without a precise description the skill either won't fire when you need it or will fire on
requests it wasn't meant for — be specific rather than writing "helps with branding".
Step 2 — hard-wire the right order of steps
Order affects the quality of the result: the symbol and slogan warm up the raw material (niche, audience, character), while the idea and focus condense it into a single thought.
Step 1: symbol — a visual anchor that can't be confused with competitors.
Step 2: slogan — a short phrase built from the same material.
Step 3: the unexpected niche idea — the "signature surprise".
Step 4: fold it all into a single positioning focus.
Each step inside is the same prompt as in a standalone generation: role, task, questions for the user, result check. The skill simply won't let you move to step 3 until step 2 is closed.
Step 3 — let the skill ask for what it needs
With separate prompts the user easily skips the questions and jumps to asking for the result. In a skill this rule has to be written out explicitly.
Skill rule: at each step, first ask the user the questions
for that step one at a time and wait for the answers. Don't move on
to the result until the inputs are gathered. Don't skip questions,
even if the user is pushing for a finished answer.
That's the one rule that actually saves time: without it Claude produces a generic symbol or slogan off a single line of context, and you have to redo the result.
Step 4 — add a stress test at the end
A good skill doesn't stop at "here's your brand" — it checks the result against a criterion shared by all four elements.
At the end of each step the skill asks: would the result pass the
unprompted-recognition test — would an outsider understand whose brand
this is, without the name and the logo. If not — the skill
offers 2 alternative options instead of one final one.
How to assemble the whole file
The finished file is the description from step 1, the four steps from step 2 written out as
ordered instructions, the questions rule from step 3 and the check rule from step 4, joined into one
coherent instruction text below the front matter. Save it as lichnyj-brend/SKILL.md and load it
like any other skill — invoked by a phrase like "build me a personal brand" or by naming it
explicitly.
How we do it
We build and test skills like these across different niches before publishing the finished file in the knowledge base — the order of steps and the questions have to actually work, not merely look logical on paper. The same principle — a process packaged into a reusable skill rather than a one-off request — underpins how the qvib engine generates code and sites step by step.
Related links
- The ready personal-brand skill and all 4 prompts: qvib.pro/arsenal/skills/lichnyj-brend
- The personal-brand prompts individually: qvib.pro/arsenal/prompts/personal-brand
- The "Claude skills" section: qvib.pro/arsenal/skills/