marketing-skills

skillcraft.cloud/marketing-skills — the shelf, the demos and the install, on one page

Marketing skills for Claude Code. For high-performing marketers going AI-native (or trying to). Type a slash command, or describe the problem in your own words, and the matching skill asks you a couple of questions, reads your files only when you say yes, and does the work: a 500-page SEO audit, a content strategy with a quarterly plan, a conversion review of the page that isn't converting, a launch plan, a LinkedIn Ads budget, a draft in your voice with the AI tells scrubbed out.

Every skill here runs real marketing work every week. This is not a showcase of skill ideas; it's the toolshed, with the sawdust left in. It's also a live list, curated by me: every few weeks I add new skills and rework or drop old ones.

What you can ask

Nothing here needs an API key to start. Connect Google Analytics and Search Console and the data-driven skills switch from asking you for numbers to pulling them.

49 skills, 4 categories. Each skill is one folder with a SKILL.md; the folder name is the slash command.

✍️ Writing — skills/Writing/

🔎 SEO — skills/SEO/

Two sources side by side: the vendored claude-seo suite by AgriciDaniel (v2.2.5, MIT, 31 skills in skills/SEO/claude-seo/) and one home-grown report skill.

/seo and /seo-audit delegate to 29 specialist skills and 18 sub-agents: technical, schema, sitemaps, hreflang, images, content quality, GEO for AI search, local and maps, ecommerce, backlinks, and connectors for Google APIs, DataForSEO, Firecrawl, Ahrefs, Bing, and more. Every specialist is also callable on its own (/seo-technical, /seo-schema, /seo-geo, ...) when you want one check without the full fan-out. Full list with one-liners: INVENTORY.md or the upstream README. The sub-agents live in skills/SEO/claude-seo/agents/ and install.ps1 copies them into ~/.claude/agents; the suite's Python runtime is created once with claude-seo setup.

⚙️ Marketing Ops — skills/Marketing-Ops/

Strategy and decisions

Research

Conversion

Reporting (HubSpot + GA4; company and CRM ids come from a per-client context pack, never from the skill)

Paid

🎨 Visual — skills/Visual/

What's in the repo

A skill folder holds its SKILL.md, and a references/ folder when the skill needs depth it
shouldn't carry in its prompt.

marketing-skills/
├── skills/
│   ├── Writing/
│   │   ├── copy-deslop/            SKILL.md
│   │   ├── copy-longform/          SKILL.md + references/
│   │   └── writing-assistant/      SKILL.md + references/
│   ├── SEO/
│   │   ├── claude-seo/             31 skills vendored from AgriciDaniel/claude-seo
│   │   │   ├── seo/                the router, plus scripts, schema and the Python runtime
│   │   │   ├── seo-audit/          the orchestrator
│   │   │   ├── seo-technical/  seo-geo/  seo-schema/  …27 more specialists
│   │   │   └── agents/             18 sub-agents, copied to ~/.claude/agents by the installer
│   │   └── seo-report/             SKILL.md + references/
│   ├── Marketing-Ops/
│       ├── content-strategy/           cro/                  customer-research/
│       ├── lead-magnets/               linkedin-ads/         marketing-council/
│       ├── marketing-monthly/          campaign-report/      product-launch/
│       └── pipeline-analysis/          SKILL.md + definitions.md + compute_funnel.py + tests + references/
│   └── Visual/
│       ├── beautify-github-readme/     SKILL.md + references/ + scripts/ (SVG render, motion GIF)
│       ├── cinemagraph/                SKILL.md + scripts/ (probe, phase renderer) + references/ + assets/template.html
│       └── frontend-design-anti-slop/  SKILL.md + references/
├── assets/readme/                  hero, section headers, motion panels, social preview
│   └── archive/                    superseded heroes with their sources
├── brand/                          the visual system
│   └── source/                     generators for every asset above
├── docs/
│   └── data-connections.md         GA4 + Search Console MCP setup
├── install.ps1                     link skills into ~/.claude/skills (Windows)
├── install.sh                      the same, on macOS and Linux
├── INVENTORY.md                    per-skill source, license and provenance
├── LICENSE                         MIT
└── NOTICE                          which license covers which folder

Company context, every skill the same way. Skills that need to know about your company (/content-strategy, /linkedin-ads, /seo-report, /cro, /customer-research, /lead-magnets, /marketing-council, and the three reporting skills) never assume a folder layout. They scan the folder you launched from, list what they found, and ask: use these, point me elsewhere, paste the context, or start with none. Nothing is read until you say so. Machine-local *.local.md files are an optional extra and never ship.

Claude Code discovers skills one folder deep under ~/.claude/skills. This repo nests them by category, so the installer links each skill into place.

The whole library

git clone https://github.com/risukisu/marketing-skills D:\marketing-skills
D:\marketing-skills\install.ps1          # junctions into ~\.claude\skills, agents into ~\.claude\agents

git clone https://github.com/risukisu/marketing-skills ~/marketing-skills
~/marketing-skills/install.sh            # symlinks into ~/.claude/skills, agents into ~/.claude/agents

Pass -Copy / --copy for plain copies instead of links, -Uninstall / --uninstall to remove them. Re-run after pulling to pick up new skills. For the SEO suite, run "$HOME/.claude/skills/seo/bin/claude-seo" setup once to create its Python runtime.

One skill: copy its folder; each skill is self-contained.

Copy-Item -Recurse .\skills\Writing\copy-deslop "$env:USERPROFILE\.claude\skills\copy-deslop"

Plugin-based skills (Superpowers, Vercel) are not in this repo; they install through the plugin marketplace.

Live data, optional. /seo-report, /seo-google, and the data hooks in /content-strategy and /cro read Google Analytics 4 and Search Console through two local MCP servers: Google's analytics-mcp and Suganthan Mohanadasan's GSC MCP. Neither is vendored here. Install steps, .mcp.json snippets, tested versions, and the two-accounts note are in docs/data-connections.md. Every skill still works without them; it asks for numbers instead.

Not everything on the shelf was built here, and the shelf says so.

Frameworks credited inline where a skill encodes someone's method: Andy Crestodina / Orbit Media (/content-strategy), AdConversion (/linkedin-ads), Luke Harries (/product-launch), Corey Haines (/cro, /customer-research, /lead-magnets, /marketing-council).

Per-skill source, license text and provenance notes live in INVENTORY.md.

Skills are plain text files, and plain text files get copied around and forgotten. This library is built so that can't happen to you:

1. One place. Every skill lives in this repo, in a folder you can read and diff. The installer links each one into ~/.claude/skills, so the file Claude Code runs is the file in the repo.
2. Updates are a pull. git pull, re-run the installer, done. No re-copying, no "which version do I have".
3. Your edits survive. Change a skill through its link and the change lands in the repo, versioned. Fork it and your fork is the library.

_{MIT for the skills written here (see LICENSE). Vendored skills keep their upstream licenses; NOTICE has the per-folder map. ☕}