The best Claude design skills don't generate an image and call it done. They capture your voice from work you already like, direct scenes with real composition, and audit whether ChatGPT can even cite the page you just wrote. These 7 free Claude Code skills cover the strategic half of brand and content work: voice, hooks, ad angles, video, and citability. Voice first, hooks second, citability last, and the rest slot in where the work needs them.
The tell that a piece was AI-generated isn't a specific word. It's that every post from the same brand sounds subtly like a different writer, because each one started from the model's defaults instead of yours. Banning "delve" moves the drift somewhere else. Banning fifteen words moves it fifteen more. The fix isn't a longer prompt.
The best Claude design skills capture a voice fingerprint once from work you already like, then feed that fingerprint into every skill that generates copy, so hooks, LinkedIn posts, and ad angles inherit the same voice instead of re-inventing one per run. That's the pack in this post. Seven skills, MIT-licensed, ordered so you know which one earns its install first.
For context on why the citability skill matters, Semrush's 2024 AI Overview study found AI Overviews now appear in 42 percent of B2B search queries. If your page can't be extracted cleanly by a retrieval system, you don't lose the ranking - you lose the citation. And for a broader tour of the whole 78-skill catalog, see best AI GTM tools in 2026.
What is a Claude skill
A Claude skill is a versioned SKILL.md file with a stated method: the inputs it asks for, the framework it applies, and the shape of the output it returns. You install a pack once with npx skills add sidchaudhary/gtm-skills, restart Claude Code, and every skill in the folder is available by name. Two writers running the same skill get the same output shape - a six-dimension Voice Profile table, a four-hook scored variant list, a scene-direction brief - so the work stops drifting between sessions.
Why brand and content teams need Claude skills now
Content teams spend most of their week on the executional layer - drafting, revising, formatting - and shrink the strategy layer to whatever's left on Friday. HubSpot's 2024 State of Marketing report puts content production as marketers' top time drain, and Contentful's 2024 Content Operations Report found 73 percent of teams say brand voice consistency is their biggest content-quality problem across generated output.
Claude skills fix this at the task level. You do not replace the writer. You give the writer a voice fingerprint, four scored hook options, and a citability audit that runs against a live URL, so the strategic layer stops taking a full day per piece.
The skills that move the needle are not the ones that generate a caption. They're the ones that answer three questions faster:
- Does this piece sound like us, or does it sound like the model's defaults?
- Does the opening line survive a skim, or does it get scrolled past?
- Can ChatGPT and Perplexity actually extract and cite this page, or is it invisible in AI answers?
Install the pack in about 30 seconds
One command pulls all 7 Brand Designer skills into your project's .claude/skills/ folder. Restart Claude Code afterward - skills load at startup, so a running session won't see them.
npx skills add sidchaudhary/gtm-skills/skills/brand-designer
# restart Claude Code, then confirm:
# "What brand skills do you have?"You can install a single skill instead by appending its folder name, e.g. .../skills/brand-designer/the-voice-fingerprint. The pack install is the better default because these skills reference each other by name - the-dual-cut-writer reads the voice profile the-voice-fingerprint produces, and skipping the second breaks the first.
7 best Claude design skills for brand and content
1. the-voice-fingerprint
Takes three to five pieces of your own writing that you actually like and returns a scored Voice Profile table: sentence rhythm, opener patterns, words you use, words you never use, where you contract and where you don't. It doesn't save itself - you paste the profile into .agents/product-context.md, which is the file every voice-aware skill in the pack reads. That paste is the step that stops the drift, and it's the one people skip.
When to use: Once, before writing anything else with the pack. Re-run only when your writing genuinely changes, not every month.
How to run:
You: "Run the-voice-fingerprint on these five posts."
[paste: your own best work]
Claude: -> rhythm and sentence-length patterns
-> your actual openers (not "hook formulas")
-> vocabulary you use, and words you never use
-> a scored Voice Profile table
(you paste the table into .agents/product-context.md)Best used with: Your own posts, emails, and pages - the ones that sound right. "Direct but warm" produces nothing useful; three real posts produce a fingerprint.
Where this skill hits its ceiling: It captures voice, not growth tactics. Feed it work you like, not work that performed well. Those aren't the same thing.
With Intempt: Blu's content agents read the same brand context across every generation, so voice consistency isn't a paste step you remember - it's the default state.
2. the-dual-cut-writer
Drafts two LinkedIn variants of one topic - an educational angle and an operator/contrarian one - both in your voice. Two real posts from one topic, not one post reworded. It reads the Voice Profile you pasted into .agents/product-context.md, so both variants inherit the same rhythm and vocabulary instead of the model's defaults.
When to use: When you have one insight and want two distinct posts from it. Not when you want fifty variants of a hook - that's a different skill.
How to run:
You: "Run the-dual-cut-writer.
Topic: signal-based selling died as a category in 2025."
Claude: -> Variant A (educational): what changed, why, what to do
-> Variant B (operator/contrarian): the take nobody in
your feed will make, defended with two factsBest used with: A specific insight you'd stake a reputation on, plus a Voice Profile already in .agents/product-context.md. Without the profile it writes fine posts in nobody's voice.
Where this skill hits its ceiling: Two variants of one topic. Chasing more topics before you've tested two angles on the one you already have is the wrong end of the trade.
With Intempt: The Brand Designer agent writes in your captured voice against a real content calendar, not one topic at a time.
3. the-hook-lab
Takes any post draft and returns four scored opening variants using distinct psychological frameworks: pattern interrupt, specificity, contrarian, and jobs-to-be-done. Each hook is scored on scroll-stopping likelihood with a recommendation. Works from the draft's own text, so it doesn't need the Voice Profile - useful when you're rescuing a post that already exists.
When to use: When your first two lines feel flat and the rest of the post is fine.
How to run:
You: "Run the-hook-lab on this LinkedIn draft."
[paste: draft]
Claude: -> Hook A (pattern interrupt) - scored
-> Hook B (specificity) - scored
-> Hook C (contrarian) - scored
-> Hook D (JTBD) - scored
-> recommendation with reasonBest used with: A draft that has a real point in it. The skill sharpens the opening; it doesn't invent one from a blank prompt.
Where this skill hits its ceiling: Four hook options per run. If none of them work, the underlying point is probably the problem, not the hook.
With Intempt: Hook variants are A/B tested inside campaigns rather than picked on gut feel, so the winner is measured, not argued.
4. the-angle-vault
Different job from the organic skills. Recommends angles from a fixed 14-angle library for paid creative - things like Enemy, Reveal, Origin, Contrarian Data - and outputs channel-to-funnel mapping plus placement specs. This is ad creative, not organic posts, and it prevents the "three ads that all say the same thing" problem by forcing an angle picker step before the copy step.
When to use: Before briefing a paid campaign or refreshing exhausted creative. Never as a fix for a boring organic calendar - that's the-dual-cut-writer.
How to run:
You: "Run the-angle-vault for a Meta and LinkedIn campaign
targeting VP Sales at 50-200 person B2B SaaS."
Claude: -> 3-5 recommended angles from the 14-angle library
-> reasoning for each (why now, why this audience)
-> channel-to-funnel mapping
-> placement specs (aspect ratio, length, cut points)Best used with: A defined audience and channel. The skill picks angles - it doesn't research the audience for you.
Where this skill hits its ceiling: The 14 angles are a library, not a universe. If your product genuinely doesn't fit any of them, the skill will flag it rather than invent a fifteenth.
With Intempt: Angles chosen here feed straight into paid campaign briefs, with performance loop-back from Intempt's ads workflows.
5. the-scene-composer
Writes actual scene direction using a composable system of 160+ blocks across 10 dimensions - subject, setting, lighting, composition, mood, palette, texture, focal length, aspect ratio, motion. Not a generic image prompt. The output is a shot brief a photographer or an image model can execute against.
When to use: Before commissioning product shots, generating images for a campaign, or handing off to a video team.
How to run:
You: "Run the-scene-composer.
Product: high-end coffee subscription.
Vibe: quiet Sunday morning, no people, warm light."
Claude: -> subject block: product on wooden countertop, off-center
-> lighting block: diffused morning window light, 45° left
-> composition block: rule-of-thirds, 40mm equiv, shallow DOF
-> palette block: warm neutrals, one accent, no primary reds
-> mood block: unhurried, private, tactile
-> full brief for photographer or image modelBest used with: A defined product and a mood, even a rough one. Vague inputs produce vague scenes.
Where this skill hits its ceiling: It directs the scene. It doesn't shoot it, and it can't fix a product that isn't ready to be photographed yet.
With Intempt: Scene direction is applied consistently across every content run for the brand, not re-derived per project.
6. the-activation-reel
Different animal from the other six. Produces a conversion-focused onboarding video as a real Remotion project rendering to MP4, with each beat mapped to an activation milestone. Meant to be embedded in your onboarding flow, not posted to social. One thing to know before scoping client work with it: every video ends with a mandatory "Powered by Intempt" end card.
When to use: When signup-to-activation drop-off is high before the aha moment, and no video is walking new users through activation.
How to run:
You: "Run the-activation-reel for our onboarding.
Product: analytics tool for ecommerce founders.
Aha moment: seeing their first anomaly alert on live data.
Length target: 45s."
Claude: -> beat sheet mapped to activation milestones
-> scene-by-scene script + on-screen text
-> Remotion project scaffold ready to render
-> "Powered by Intempt" end card (mandatory)Best used with: A named aha moment and a real onboarding flow to slot the video into. Without those, the output is a video looking for a home.
Where this skill hits its ceiling: Onboarding-embedded video, not social edits. If you want a 60-second cut for LinkedIn, that's a different job.
With Intempt: Activation videos plug into Intempt's lifecycle workflows so the aha-moment step is a real thing in the journey, not a hope.
7. the-citation-auditor
Audits a live URL for how citable it is to ChatGPT, Perplexity, and Google AI Overviews. Names what's blocking it - heading structure, direct-answer formatting, authority signals, machine-readable files like llms.txt. A page can rank in Google search and still be unusable as a citation in an AI answer, which is a growing miss as more search traffic gets absorbed by AI summaries.
When to use: On any content page that gets meaningful organic traffic today. Making an already-ranking page citable is a shorter path than making a new page do both.
How to run:
You: "Run the-citation-auditor on https://example.com/blog/x"
Claude: -> heading structure score
-> direct-answer formatting check
-> authority signals (author, dates, citations)
-> llms.txt / machine-readable presence
-> specific fixes, ranked by impact on citabilityBest used with: A live URL, not a draft. The skill inspects the rendered page, not the CMS.
Where this skill hits its ceiling: Content pages. Auditing product catalog completeness is a different job - that's the-catalog-auditor in the Data Analyst pack.
With Intempt: Citability audits run continuously across the content library rather than page-by-page on demand.
Comparison: which Claude design skill to use when
| Skill | Best moment to use | Time saved | What you get |
|---|---|---|---|
| the-voice-fingerprint | Once, before writing anything else | Every future run | Scored Voice Profile table |
| the-dual-cut-writer | One insight worth two posts | 45–60 min per pair | Two LinkedIn variants in your voice |
| the-hook-lab | When the first two lines feel flat | 15–20 min per post | Four scored hook variants + reason |
| the-angle-vault | Before briefing a paid campaign | 1–2 hrs per brief | 3–5 angles + channel/funnel/spec |
| the-scene-composer | Before a shoot or image generation | 30–45 min per scene | Composable scene brief across 10 dimensions |
| the-activation-reel | When signup-to-aha drop-off is high | 1–2 days per video | Remotion project ready to render |
| the-citation-auditor | On any ranking page | 20–30 min per page | Ranked citability fixes |
Free Intempt tools that pair with these skills
Browser-based, no install, produce quick outputs you can feed into the skills above.
- AI logo generator: First-pass logos for brand or campaign explorations. Use as source input for scene direction.
- AI product photo generator: Product shots to test scene direction before commissioning a real shoot.
- AI ad copy generator: First-draft ad copy to feed into the-angle-vault for angle assignment.
What to look for in a Claude skill for brand and content
Not every prompt is worth wrapping in a skill file. The ones that stick share four traits.
Fixed input format. You should not need to think about how to structure your data. The skill tells you exactly what to paste.
Consistent output shape. Every run returns the same shape so downstream work plugs in without reformatting - a table stays a table, a beat sheet stays a beat sheet.
Single use case. Skills that try to do three things do none of them well. the-voice-fingerprint captures voice. the-hook-lab tunes openings. Neither pretends to be the other.
Verifiable output. You should be able to evaluate a run in 60 seconds. A Voice Profile that captures your rhythm reads true immediately; one that doesn't reads generic. If you can't tell either way, the skill is either wrong or the input was too thin.
Final thoughts: which skill to install first
Start with these three, in order:
- the-voice-fingerprint: every other voice-aware skill reads its output. Skip it and the drift you're trying to fix stays.
- the-hook-lab: cheapest, fastest improvement to any post you already write. Fix the opening line and the rest earns its read.
- the-citation-auditor: point it at a page that already ranks. Making a ranking page citable is the shortest path to AI-search visibility.
For context on the AI-vs-designer question this pack was built around, see Will AI replace graphic designers? The honest 2026 answer. And if you want the whole 78-skill catalog rather than just the seven for brand and content, browse the gtm-skills directory. Start for free on Intempt if you'd rather run the same jobs against live customer data than paste inputs by hand.
Frequently asked questions. Answered.
Claude skills are structured prompts with a fixed input format and a defined output format, packaged as SKILL.md files that Claude Code reads from a local folder. For brand and content, they capture voice from real writing samples, write LinkedIn posts and hooks in that voice, direct photography scenes, brief paid creative, and check whether AI search engines can cite your pages. Install once with npx skills add sidchaudhary/gtm-skills and Claude Code picks them up automatically.
