How skills work
A skill is an expert playbook Claude follows. You don’t configure anything, you just ask, and Claude runs the right one.
Two ways to run any skill
Section titled “Two ways to run any skill”1. Just ask, in plain English. Describe what you want and Claude picks the skill.
“Research the cold-plunge market for examplebrand.com”
“Run the Monday routine on examplebrand”
“Why did CPA spike overnight?”
2. Call it by name. Every skill has a short name (like apollo or hermes). On the desktop app you can type it as a command with the toolkit’s prefix:
/um-toolkit:apollo
Both work. Plain English is usually faster; calling by name is handy when you know exactly which skill you want.
Every skill reads your brand files first
Section titled “Every skill reads your brand files first”Before a skill produces anything, it loads your brand files: your offer, your voice, your buyer personas, your objections. That’s what makes the output sound like your brand. If you haven’t onboarded your brand yet, the skill will route you to setup first.
Skills tell you when they can’t do something
Section titled “Skills tell you when they can’t do something”This toolkit is unusually blunt about gaps, and that is deliberate. A check that could not run comes back as NO-DATA, and NO-DATA blocks exactly like a failure does. You will see blocked launches, restricted sends, and rows that say “not ranked, this input is missing” rather than confident numbers built on guesses.
Worth ten minutes before you rely on any verdict: gates and verdicts.
How to read these pages
Section titled “How to read these pages”Every skill has its own page, laid out the same way:
- What it does, in plain sentences.
- When to reach for it: the situations it’s built for.
- Where it runs: Cowork, the desktop app, or both.
- How to run it: the exact words to type.
- What it needs from you, and what you get back.
- A worked example, and tips & gotchas.
Runtime tags
Section titled “Runtime tags”- Cowork-ready: works in both Cowork and the desktop app.
- Desktop: needs the Claude Code desktop app. These render images or video, or build and deploy pages, which Cowork can’t do yet. See Choose your setup.
The full catalog
Section titled “The full catalog”44 skills plus 2 bundled tools.
Setup and onboarding
Section titled “Setup and onboarding”| Skill | What it’s for | Where it runs |
|---|---|---|
| getting-started | Onboards you and your brand on a fresh install | Cowork-ready |
| brand-context | Researches a URL and fills in all your brand files | Cowork-ready |
| onboard-client | Adds another brand or client to your workspace | Cowork-ready |
| client-brain | Builds the 29-module Brain: research plus intake | Cowork-ready |
| business-brain-builder | Brain 2.0: connected accounts, 24 months of history, self-updating | Cowork-ready |
Run the week
Section titled “Run the week”| Skill | What it’s for | Where it runs |
|---|---|---|
| weekly-account-os | The Monday button. Runs the whole weekly cadence, one action list | Cowork-ready |
| weekly-readout | The client-facing report, wins and losses both stated | Cowork-ready |
| meta-account-doctor | Full Meta account audit, leak-ranked | Cowork-ready |
| creative-testing-os | Scale / iterate / kill board with the spend floor shown | Cowork-ready |
Diagnostics
Section titled “Diagnostics”| Skill | What it’s for | Where it runs |
|---|---|---|
| tracking-doctor | Can I trust my data? Six layers, one trust grade | Cowork-ready |
| signal-surgeon | The Meta signal layer: EMQ, dedup, exact payload fixes | Cowork-ready |
| drop-triage | Why is performance down? Scores all four causes | Cowork-ready |
| deliverability-doctor | Are we landing in spam? Plus the send verdict | Cowork-ready |
Launch and scale
Section titled “Launch and scale”| Skill | What it’s for | Where it runs |
|---|---|---|
| new-client-launch | Signed to live traffic, behind a three-legged hard gate | Cowork + Desktop |
| meta-launch-architect | The launch spec, with the budget arithmetic shown | Cowork-ready |
| meta-scaling-copilot | Scale, hold or cut. Per-diff approval, nothing auto-writes | Cowork-ready |
| policy-compliance-gate | A policy verdict before the spend, not after the ban | Cowork-ready |
Message and offer
Section titled “Message and offer”| Skill | What it’s for | Where it runs |
|---|---|---|
| offer-architect | Scores the offer before you spend behind it | Cowork-ready |
| voc-miner | What customers actually say, traced to real verbatims | Cowork-ready |
| angle-architect | The angle map everything else executes against | Cowork-ready |
| hook-matrix-forge | 7 to 15 openings per angle, one variable per cell | Cowork-ready |
| concept-lab | Named creative worlds, sized to your spend | Cowork-ready |
| marketing-psychologist | Psych audit and lever stack, evidence-graded | Cowork-ready |
| retention-architect | Flow map, every gap priced, 8 to 12 build specs | Cowork-ready |
Research and strategy
Section titled “Research and strategy”| Skill | What it’s for | Where it runs |
|---|---|---|
| apollo | 10-module market and customer research | Cowork-ready |
| persona-architect | 5 deep buyer personas with psychology | Cowork-ready |
| brand-style-extractor | Pulls a brand’s real colors, fonts and logo | Desktop |
| ad-forensic-teardown | Systematic competitor-ad teardown | Cowork-ready |
Copywriting
Section titled “Copywriting”| Skill | What it’s for | Where it runs |
|---|---|---|
| hermes | Direct-response ads, emails and sales copy | Cowork-ready |
| legendary-sales-letter | Long-form sales letters and advertorials | Cowork-ready |
| conversion-copy-architect | Behavioral-science conversion copy | Cowork-ready |
| ad-overlay-copy-engine | Text-overlay copy and layout for ad images | Cowork-ready |
Ad creative
Section titled “Ad creative”| Skill | What it’s for | Where it runs |
|---|---|---|
| meta-image-ad-generator | Meta image-ad concepts and prompts | Desktop (to render) |
| seedance-video-ad-generator | Realistic AI video ad prompts for Reels, TikTok, Shorts | Desktop (to render) |
Video (not AI footage)
Section titled “Video (not AI footage)”Four skills, one job each. Start with motion-video, it picks the stack for you.
| Skill | What it’s for | Where it runs |
|---|---|---|
| motion-video | The entry point. Reads the brief, picks the stack, hands over | Desktop |
| hyperframes | The HTML/CSS/GSAP stack, complete. The default | Desktop |
| remotion | React/JSX: per-word captions, 10+ variants, springs, 3D | Desktop |
| website-to-video | URL in, video out | Desktop |
Pages and funnels
Section titled “Pages and funnels”| Skill | What it’s for | Where it runs |
|---|---|---|
| page-architect | Full page build: research, design, build, QA, ship | Desktop |
| design-architect | Design system, mockups and visual QA | Desktop |
| living-elements | Animated “living” hero assets for premium pages | Desktop |
| page-craft-engine | The shared craft engine inside page builds | Not user-invocable |
Review and QA
Section titled “Review and QA”| Skill | What it’s for | Where it runs |
|---|---|---|
| red-team | Adversarial “would my buyer actually convert?” review | Cowork-ready |
| unicorn-page-review-panel | World-class expert page scoring | Cowork-ready |
Bundled tools
Section titled “Bundled tools”Other skills call these. They aren’t slash-skills you invoke directly.
| Tool | What it’s for | Where it runs |
|---|---|---|
| creatify-batch | Batch image generation via Creatify | Desktop |
| image-sourcer | On-demand image generation | Desktop |
Where to start
Section titled “Where to start”- Brand new? getting-started, then brand-context.
- Managing an account weekly? weekly-account-os is the one command that runs the routine.
- New client just signed? new-client-launch covers signed to live.
- Something broke? drop-triage for a sudden drop, tracking-doctor if you suspect the numbers.