A skill that makes Claude audit a brand you already have, the way a senior design director would, across the site, social, packaging, ads, email and shops. It starts from the decision the audit feeds, such as whether to rebrand before a retail pitch, then logs every touchpoint with dated evidence and counts the versions of each element in use. It scores every item on a fixed rubric for consistency and strength, compares three to five competitors on the same scale, ranks the top problems by the money that passes through them against the cost to fix, and ends with a 30/60/90-day plan sized to your team's hours.
Inside are five reference files. An inventory kit lists every touchpoint by area, with tiers, sample sizes, capture specs and a way to sample colours from screenshots in code. The scoring rubric gives a written anchor for each score from 0 to 4, plus two pass or fail gates for accessibility and truth. A rules file quotes WCAG 2.2, the FTC's endorsement and Made in USA rules, CAN-SPAM, Gmail sender rules, trademark symbols and food label type sizes, each sourced and checked on 17 October 2026. One file covers competitors, cost and the plan; another is a full audit of a fictional hot sauce brand. It never invents a number. It leaves a marked gap and asks.
Claude cannot judge printed colour from a phone photo, see a logged-in page or stand in front of a shelf, so you capture those, and it reads live sites only when web search is on. With code execution on, it samples colours, measures contrast and hands back the scorecard as a spreadsheet and the report as a deck. The rules file covers the US, and it flags legal questions for your lawyer instead of answering them. You get the most from it by bringing your guidelines, full-page screenshots at two widths, daylight pack photos and your monthly numbers. Without revenue and spend figures, it can only rank problems by how many people see them.
Claude skill · version 1.0
Brand audit
Checking your access.
In the download · 7 files
- SKILL.mdThe method Claude follows: scope, inventory, scorecard, competitors, problems ranked by cost, then the 30/60/90 plan and the report, one stage at a time.
- README.mdHow to install it in claude.ai or Claude Code, how to start, and the licence.
- references/inventory-kit.mdThe touchpoint list across site, social, packaging, ads, email and stores, with tiers, sample sizes, the evidence request, capture specs, the inventory log and colour sampling in code.
- references/scoring-rubric.mdEight criteria with a written anchor for every score from 0 to 4, the accessibility and truth gates, weights, the consistency and strength map and the scorecard template.
- references/rules-and-specs.mdWCAG 2.2, FTC endorsement, review and Made in USA rules, CAN-SPAM, Gmail sender rules, inbox logos, trademark symbols, food label type sizes, favicons, link previews and ad libraries, sourced and checked on 17 October 2026.
- references/competitors-and-cost.mdChoosing three to five competitors, the comparison table, category codes and distinctive assets, money at stake, break-even lift, the problem card and the rules for the 30/60/90 plan.
Install it and begin
01Install it in Claude CodeTerminal mkdir -p ~/.claude/skillsunzip ~/Downloads/brand-audit.zip -d ~/.claude/skills/# Start it with /brand-audit. 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 audit skill. Audit [BRAND NAME], which sells [WHAT YOU SELL] in [COUNTRY]. The audit is for [THE DECISION, FOR EXAMPLE REBRAND OR NOT]. Cover [SITE, SOCIAL, PACKAGING, ADS, EMAIL, STORES]. Our competitors are [2 TO 4 COMPETITOR NAMES]. Start with the scope and tell me what to capture.Attach your brand guidelines if you have them, full-page screenshots at desktop and phone width, and daylight photos of every pack. Add monthly revenue, ad spend and your design rate if you want problems ranked by cost.
Side by side
What goes in each part of the 30/60/90 plan
| Days 1 to 30 | Days 31 to 60 | Days 61 to 90 | |
|---|---|---|---|
| Its job | Stop the damage and lock the core | Make the fixes stick | Make the bigger changes, then measure |
| What goes in | Every failed gate, and every fix under 4 hours on the most-seen touchpoints | Templates for anything made more than twice a month, the freelancer brief and marketplace listings | Changes timed to a print run, a site sprint or a shoot, then a re-score |
| Typical fixes | A false badge removed, a button darkened to pass 4.5:1 contrast, one folder of final logo files | Social and email templates, a one-page rules sheet, a marketplace listing moved onto the current logo | New labels at the next print run, a rebuilt landing page, ads that carry the brand's own cue |
| Packaging | Sticker or relabel only for a legal error | Redesign the label in time for the next run | Print the new label at the next run |
| Fix level | Mostly the single instance, plus anything legal | The template and the rule | The rule and the system: who owns and approves the work |
| Done when | No gate fails and the core fits on one page | A freelancer can make an on-brand post without asking | The most-seen touchpoints are re-scored on the same rubric and the change is written down |
Legal and accessibility failures go first, whatever they cost to fix. Cosmetic packaging changes wait for the next print run. When the hours do not fit, the plan is cut from the bottom of the ranked list.
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.
- Wayback MachineFree
Shows what a brand's site looked like at any point in its history.
Bad at Captures are patchy and heavy pages often archived broken.
- BrandfetchFree tier, paid above it
Pulls a company's logo, colours and typeface from its public site through an API.
Bad at It reports what a site uses, which is often not what the guidelines say. Treat it as evidence, not as truth.
Tells you whether two colours are legible together, against the published standard.
Bad at The ratio is a floor and not a design. Passing text can still be unpleasant to read.
Why it matters
Ask a chat to audit a brand and you get opinions about the logo. Nothing is counted, nothing is scored the same way twice, and a false Made in USA badge sits in the same list as a font choice. This skill makes Claude log the evidence first, score it against your own rules, treat legal and accessibility failures as stops, and rank the rest by money at stake, so the 90-day plan fixes what costs you most.
- Type
- technique
- Period
- ongoing