Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
<div align="center">
<img src="docs/rampstack-complete-banner.jpg" alt="Complete Claude Skills for the full web lifecycle. Build, ship, audit, optimize." />
Brand Build Skills for Claude
A complete, opinionated library of [Claude Skills](https://docs.claude.com/en/docs/agents-and-tools/agent-skills/overview) covering the full lifecycle of building, launching, running, and growing a brand and a website.
</div>
<!-- COUNT_INTRO:START -->
107 stack-agnostic skills covering brand, design, content, SEO, dev, ops, growth, and research. Includes an Ahrefs MCP-powered SEO audit suite. Use them on Next.js, WordPress, Shopify, Webflow, plain HTML, or anything else.
<!-- COUNT_INTRO:END -->
Featured in [awesome-claude-skills](https://github.com/ComposioHQ/awesome-claude-skills) under Business & Marketing.
Install in Claude Code
Add the marketplace, then install the plugin you want:
/plugin marketplace add rampstackco/claude-skills
# full catalog
/plugin install rampstack-skills@rampstack
# focused subsets
/plugin install rampstack-starter@rampstack
/plugin install rampstack-seo@rampstack
/plugin install rampstack-pm@rampstackPrefer a lighter marketplace that lists only the curated subsets (no full catalog)? Add rampstackco/plugins instead and install the same three plugins from there:
/plugin marketplace add rampstackco/plugins
/plugin install rampstack-starter@rampstackSkills load on demand: each contributes roughly its name and description until Claude needs it.
Table of contents
- What are Claude Skills?
- What is in this library
- Featured skills
- See it in action
- Logo design in action
- Reference build in action
- Getting started
- Quick example
- How they compose
- How the catalog connects
- Surfaces
<!-- COUNT_TOC:START -->
<!-- COUNT_TOC:END -->
- Recommended MCPs
- Authoring conventions
- Repository structure
- Trust and security
- Contributing
- Acknowledgments
- Resources
- License
What are Claude Skills?
Claude Skills are reusable capability packages that teach Claude how to handle a specific kind of task with a consistent framework, vocabulary, and output format. Each skill is a folder containing a SKILL.md (instructions plus YAML metadata) and optional reference files (templates, checklists, worked examples). Claude loads a skill automatically when a user request matches the skill's description.
Skills work across Claude.ai, Claude Code, and the Anthropic API. Once you write a skill, it is portable across all three.
For the official deep dive, see Anthropic's Agent Skills documentation.
What is in this library
This is not a curated list of other people's skills. It is a single, opinionated library where every skill follows the same structure and conventions, so the skills compose cleanly across a real project lifecycle.
What you get:
<!-- COUNT_WHATYOUGET:START -->
- 107 skills across 16 categories, every one with a complete
SKILL.mdand at least one reference file - 516 reference files (templates, checklists, decision matrices, worked examples)
<!-- COUNT_WHATYOUGET:END -->
- Stack-agnostic. Works on any web stack. The only named-tool exception is the SEO audit suite, which assumes the Ahrefs MCP.
- Future-proof. Principles over tools. Stable concepts over trending techniques. References to durable specs (W3C, WHATWG, Schema.org, MDN, NN/g, WCAG) over content that ages with each algorithm update.
- Uniform structure. Every skill uses the same section order, the same tone, and the same authoring conventions. Predictable in, predictable out.
- Composable. Skills reference each other.
creative-briefpoints tobrand-voice.incident-responsepoints tomonitoring-and-alerting. Each skill's "When NOT to use" tells you which sibling fits your adjacent work.
Highlight categories: brand strategy and identity, design systems, content production with full Tier 1 and Tier 2 coverage, full SEO suite (foundation plus Ahrefs MCP-powered audit suite), product management with experimentation and gap-closing tracks, growth tooling for interactive web tools, paid media discipline, frontend dev and accessibility, performance and QA, launch and incident ops, UX research, plus a meta-skill that teaches you to write your own.
Featured skills
<!-- FEATURED_SKILLS:START --> Six entry-point skills, one per audience track. Run any of these standalone, or compose them with the rest of the catalog.
| Skill | What it does |
|---|---|
creative-direction (Brand and creative) | Four-axis brief (tone, aesthetic, audience, sensory ambition) that gives every downstream skill a coherent direction |
experiment-design (PM, experimentation) | From hypothesis to decision: sample size, duration, segment analysis, and the failure modes that produce wrong shipping calls |
feature-launch-playbook (PM, gap-closing) | The discipline of launching a feature well: positioning, internal alignment, customer comms, enablement, rollout, monitoring |
pillar-content-architecture (Content) | Hub-and-cluster topical authority: pillar selection, cluster planning, internal linking, refresh discipline |
landing-page-copy (Marketing) | Landing pages, sales pages, hero-to-CTA flow with copy that converts |
funnel-flow-architecture (Growth tooling) | Cross-tool conversion flows architected to match the audience and the funnel stage |
<!-- FEATURED_SKILLS:END -->
See it in action
<strong><a href="https://rampstack.co/showcase/creative-direction" target="_blank" rel="noopener">The creative-direction skill rendered as a live showcase →</a></strong>
Forty-two fictional brands generated from briefs that all use the same skill. Each is a fully styled brand site, not a mockup. The showcase demonstrates what the four-axis framework produces in practice and lets you filter by axis position to see how each combination renders.
<p align="center"> <picture> <source media="(max-width: 640px)" srcset="assets/showcase/creative-direction-highlight-mobile.jpg"> <img src="assets/showcase/creative-direction-highlight-desktop.jpg" alt="Creative Direction skill highlight diagram. Navy header card reads 'Impactful Creative Direction' with the subtitle 'Direction for art, taste, and style'. Four quadrants below show: Framework Axes 4 (Tone, Aesthetic, Relationship, Sensory), Framework Positions 16 (each axis combines into 16 distinct positions), Example Treatments 42 (Pulse, Bloom, Forge, Observatory, and 38 others), and Possible Compositions infinity (Motion: Static, Light, Medium, High). Caption reads 'No templates, only guided outputs.'"> </picture> </p>
<p align="center"> <img src="assets/showcase/showcase-grid-hero.png" alt="Showcase grid of brand archetypes including Pulse, Volt, Anode, Drift, and others, with type and motion intensity filter pills above the cards." /> </p>
Filter by any axis position
The skill defines four axes: tone, aesthetic, relationship, sensory. The showcase lets you filter by any combination and see which examples match. Pre-filtered URLs deep-link from the SKILL.md and axes-explained reference, so you can read about a position and click straight through to the rendered examples.
<p align="center"> <img src="assets/showcase/showcase-filter-active.png" alt="Showcase grid filtered by Tone equals Provocative and Sensory equals Resonant, showing eight matching brand cards with the axis disclosure auto-expanded." /> </p>
The empty state is the lesson
The framework is generative. The showcase is illustrative. Most rare-but-powerful combinations are valid creative choices that simply have not been built yet. Set Provocative + Editorial Restrained + Coach + Resonant and the grid is empty.
<p align="center"> <img src="assets/showcase/showcase-empty-state.png" alt="Showcase grid with all four axis filters set to Provocative, Editorial Restrained, Coach, and Resonant, showing zero matching examples and the empty state copy: No example yet. The framework allows this combination, it just hasn't been built as one of the thirty worked examples." /> </p>
The framework's range
Same skill, same brief format. Four completely different visual systems. Notice that Pulse and Bloom share identical axis positions yet read as opposite visual languages. The reference brands and aesthetic interpretation do the rest.
<table> <tr> <td width="50%"><img src="assets/showcase/archetype-pulse.png" alt="Pulse music streaming brand. Saturated gradient hero with the headline 'Sound that moves with you' and pink-to-cyan equalizer bars below." /></td> <td width="50%"><img src="assets/showcase/archetype-forge-fitness.png" alt="Forge boutique fitness studio. Dark industrial hero with intense typography and motivational copy." /></td> </tr> <tr> <td><strong>Pulse</strong> · music streaming<br/><em>Sound that moves with you.</em><br/>Playful / Expressive Maximalist / Companion / Resonant<br/><a href="https://rampstack.co/showcase/creative-direction/pulse-music" target="_blank" rel="noopener">See Pulse demo example →</a></td> <td><strong>Forge</strong> · boutique fitness<br/><em>Show up. Get hammered.</em><br/>Provocative / Expressive Maximalist / Coach / Resonant<br/><a href="https://rampstack.co/showcase/creative-direction/forge-fitness" target="_blank" rel="noopener">See Forge demo example →</a></td> </tr> <tr> <td><img src="assets/showcase/archetype-bloom-soda.png" alt="Bloom adaptogenic soda brand. Peachy gradient hero with tri-color headline 'Soda that loves you back' and a strawberries-around-soda-can product photo." /></td> <td><img src="assets/showcase/archetype-observatory-editorial.png" alt="Observatory Editorial. Cream paper hero with restrained serif headline 'An observability tool for the engineers who already know what they are doing'." /></td> </tr> <tr> <td><strong>Bloom</strong> · adaptogenic soda<br/><em>Soda that loves you back.</em><br/>Playful / Expressive Maximalist / Companion / Resonant<br/><a href="https://rampstack.co/showcase/creative-direction/bloom-soda" target="_blank" rel="noopener">See Bloom demo example →</a></td> <td><strong>Observatory Editorial</strong> · observability tool<br/><em>An open-source tool that respects engineer time.</em><br/>Conversational / Editorial Restrained / Peer / Considered<br/><a href="https://rampstack.co/showcase/creative-direction/observatory-editorial" target="_blank" rel="noopener">See Observatory demo example →</a></td> </tr> </table>
<p align="center"> <strong><a href="https://rampstack.co/showcase/creative-direction" target="_blank" rel="noopener">See all the brands in the showcase →</a></strong> </p>
Run this on your own brand
The creative-direction skill lives at skills/creative-direction/. Install it (see below), give Claude a project name and a few inspiration references, and the skill walks you through producing a brief that downstream skills can consume. The brand sites in the showcase were built from briefs of exactly that shape.
Logo design in action
The logo-design skill is rendered on rampstack.co as two parallel surfaces. The variant explorer goes deep on one brand at a time: a primary mark, variants across architectures, applied contexts. The taxonomy gallery goes wide across the architecture space: ten fictional marks demonstrating eight mark architectures (wordmark, lockup, monogram, letterform-as-symbol, abstract, pictorial, combination, emblem). Same skill, two different lenses.
<p align="center"> <picture> <source media="(max-width: 640px)" srcset="assets/showcase/logo-design-highlight-mobile.jpg"> <img src="assets/showcase/logo-design-highlight-desktop.jpg" alt="Logo Design skill highlight diagram. Navy header card reads 'Bespoke Logo Design' with the subtitle 'Bringing brands to life.' Below the header, a simulated construction guide shows a stylised letterform B rendered against gridlines, with a Golden Ratio overlay, Primary Curve and Secondary Shape callouts, a Kerning marker, and a six-swatch color palette. Three columns at the bottom show: Verticals (Tech, Finance, Healthcare, Retail), Brand Voice (Trustworthy, Innovative, Premium, Approachable), and Architectures (Monogram, Wordmark, Emblem, Abstract)."> </picture> </p>
Per-brand depth
<strong><a href="https://rampstack.co/showcase/logo-design" target="_blank" rel="noopener">The variant explorer →</a></strong>
Each brand has a primary mark plus variants across architectures and applied contexts. The logo-design skill walks through the discipline of choosing one architecture and rendering it consistently across the system the brand will actually use.
<p align="center"> <picture> <source media="(max-width: 640px)" srcset="assets/showcase/logo-design-showcase-mobile.png"> <img src="assets/showcase/logo-design-showcase-desktop.png" alt="Logo design variant explorer showing six fictional brand cards in a three-by-two grid: Whitfield Carter (legal counsel lockup), Wren and Bough (consumer goods lockup), Highline (hospitality wordmark), Sentinel (tech and AI symbol-only), Lacuna (fashion wordmark), and Roost (restaurant lockup). Each card pairs a primary mark with three classification chips for architecture, typographic register, and category, plus a four-variants and five-application-contexts subtext."> </picture> </p>
The brands are filterable by architecture, typographic register, and category. The intent is reference work, not consumable templates.
Architectural taxonomy
<strong><a href="https://rampstack.co/showcase/logos" target="_blank" rel="noopener">The marks gallery →</a></strong>
Ten fictional marks across eight mark architectures: wordmark, lockup, monogram, letterform-as-symbol, abstract, pictorial, combination, emblem. The taxonomy makes the architectural distinctions concrete by showing all eight side-by-side, with three wordmarks at three typographic registers so the architectural label does less work than the execution.
<p align="center"> <picture> <source media="(max-width: 640px)" srcset="assets/showcase/marks-showcase-mobile.png"> <img src="assets/showcase/marks-showcase-desktop.png" alt="Marks gallery showing six fictional brand cards in a three-by-two grid: knurl (lowercase serif wordmark with knurled texture), TARSUS (uppercase sans lockup with stacked-bar mark), PLINTH (classical serif inside a double-lined emblem frame), Caval (italic horse silhouette plus italic wordmark combination), Ostend (flowing OS monogram resolving to a single connected glyph), and GLINT (high-contrast Didone wordmark with hairline I crossbar). Each card carries the brand name, descriptor, and three classification chips for mark architecture, vertical, and brand voice."> </picture> </p>
Filter by architecture, vertical, or brand voice; click any mark card to read its design rationale.
Reference build in action
The full catalog rendered as a 4-phase reference build: blank brief through deployed audited launch site. Threshold is a fictional PLG onboarding analytics product, but the research, brand foundations, build, and audit findings are all real. The reference build is the catalog's single strongest demonstration of how the skills compose end-to-end.
<p align="center"> <picture> <source media="(max-width: 640px)" srcset="assets/showcase/reference-build-highlight-mobile.png"> <img src="assets/showcase/reference-build-highlight-desktop.png" alt="Reference build hero card. Navy header card reads 'Reference Build' with subtitle 'A fictional B2B SaaS launch, end-to-end.' Below, four white phase cards arranged in a 2x2 grid: Phase 01 Strategy and Research with caption 'Real Ahrefs research applied to a fictional brief'; Phase 02 Brand and Design with caption 'Working brand system with live tokens and components'; Phase 03 Build and Ship with caption 'Deployed launch microsite at /demo/threshold'; Phase 04 Audit and Optimize with caption 'Real audit findings with applied fixes.' Footer caption reads 'Threshold is fictional. The methodology is not.'"> </picture> </p>
The four phases
[Phase 1: Strategy and research →](https://rampstack.co/walkthroughs/saas-launch-research-and-strategy) Real Ahrefs keyword research, competitor analysis, content gap audit, and backlink opportunity mapping applied to a fictional B2B SaaS brief. Live data tables sourced from the Ahrefs API.
[Phase 2: Brand and design →](https://rampstack.co/walkthroughs/saas-launch-brand-and-design) Logo system, color and typography tokens, working brand component primitives. The brand system renders live on the walkthrough page in real fonts and tokens, not just described.
[Phase 3: Build and ship →](https://rampstack.co/walkthroughs/saas-launch-build-and-ship) The actual launch microsite built with Next.js using Phase 2's brand foundations. Live at rampstack.co/demo/threshold. Persistent demonstration banner; noindex; local-only waitlist form.
[Phase 4: Audit and optimize →](https://rampstack.co/walkthroughs/saas-launch-audit-and-optimize) Real audit on the deployed site using the catalog's audit suite (axe-core, Lighthouse, manual checks). Real findings with severity, real fixes applied, real before/after metrics. The closing chapter where the catalog audits its own output.
The deployed result
The four phases compose into a working microsite at rampstack.co/demo/threshold. Real Next.js code, real brand foundations from Phase 2 referenced via cross-route imports, real working multi-step waitlist form (no data stored), persistent demonstration banner, and the inline data visualizations that came out of the post-audit polish pass.
<p align="center"> <picture> <source media="(max-width: 640px)" srcset="assets/showcase/threshold-demo-screenshot-mobile.png"> <img src="assets/showcase/threshold-demo-screenshot-desktop.png" alt="Full-page screenshot of the deployed Threshold demo at rampstack.co/demo/threshold. Persistent navy demonstration banner at top reads 'Demonstration · Threshold is a fictional product built to illustrate how the catalog composes from blank brief to deployed launch microsite.' Below, the hero section shows a serif headline 'Know how new users actually get to value' next to a stylized product dashboard mockup with KPI tiles, an activation funnel chart, and recent cohorts comparison. Further down the page: a fictional cohort trust strip, a 'The gap' problem section, a wedge section with inline funnel and time-to-first-value charts, a comparison table against Mixpanel/Amplitude/Heap and Pendo/Userpilot, a 'How it works' section with three connected cards, a multi-step waitlist form, and a FAQ section."> </picture> </p>
Why a fictional product
Real client work cannot be open-sourced; portfolio claims trigger conflict-of-interest concerns in interviews and consulting conversations. A fictional product with a documented brief plus real research, real brand foundations, real working code, and real audit findings produces a teaching artifact that demonstrates methodology without claiming relationships. Threshold is a measurement tool that does not exist; the methodology that built it is the catalog working end-to-end.
Getting started
Skills install in a different place for each platform. Pick the one that matches your workflow.
Option 1: Claude.ai (web and desktop)
If your Claude.ai plan supports custom Skills:
- Go to Settings → Capabilities → Skills.
- Upload the skill folder you want as a
.zip(one zip per skill folder containingSKILL.mdand thereferences/subfolder). - Enable the skill in the chat interface.
Claude will load the skill automatically when your request matches its description.
For current plan availability and the exact upload UI, see Anthropic's Skills user guide.
Option 2: Claude Code (recommended)
Skills are first-class citizens in Claude Code. Drop them into your skills directory and Claude Code picks them up automatically.
User-level skills (available in every project):
# macOS / Linux
mkdir -p ~/.claude/skills
cp -r skills/* ~/.claude/skills/
# Windows (PowerShell)
New-Item -ItemType Directory -Force -Path "$HOME\.claude\skills"
Copy-Item -Recurse skills\* "$HOME\.claude\skills\"Project-level skills (available only in a specific project):
mkdir -p .claude/skills
cp -r path/to/this-repo/skills/* .claude/skills/Start (or restart) Claude Code. Skills load automatically.
For exact current paths and config flags, see the Claude Code documentation.
Option 3: Anthropic API
Use Skills programmatically by referencing them in your API calls. Skills must first be uploaded to your workspace (via the Console or API), then referenced by ID when creating messages.
For the current API surface, request format, and limits, see the Agent Skills API documentation.
Option 4: pi and Codex
Prebuilt distributions install the catalog into a project's .agents/skills folder for two agents outside Claude:
- pi:
dist/pi/README.md - Codex:
dist/codex/README.md
Want only a few skills?
You do not have to install all of them. Pick the categories that match your work. The library is modular: each skill stands on its own.
Quick example
Once installed, skills trigger automatically based on your request. You do not have to name the skill or change how you talk to Claude.
You ask:
"Our organic traffic dropped 30% last week. Help me figure out why."
What happens:
Claude recognizes the request matches seo-traffic-diagnosis, loads the skill, and walks through its 5-layer root cause framework: confirm the change is real → localize the change → page-level analysis → technical analysis → external analysis. By the end, you have a hypothesis statement, evidence, and an action plan, structured the same way every time.
Other natural triggers:
- "Help me write a creative brief" →
creative-brief - "Audit my homepage for SEO" →
seo-onpage - "We need a backlink audit" →
seo-backlink-audit - "Plan our content roadmap for Q3" →
seo-content-gap-auditpluscontent-strategy - "Postmortem template for last night's incident" →
after-action-report - "How do I write my own skill?" →
skill-creation-walkthrough
You can also call a skill explicitly: "Use the seo-audit-orchestration skill to run a full audit on example.com."
How they compose
The skills compose into a full project flow:
brand-discovery → brand-ideation → brand-identity → brand-style-guide → brand-voice
↓
creative-brief → information-architecture → content-strategy → design-system
↓
seo-keyword → seo-content-audit → content-and-copy → landing-page-copy
↓
seo-onpage → seo-technical → seo-aeo-geo → seo-offpage → seo-competitor
↓
frontend-component-build → accessibility-audit → performance-optimization
↓
code-review-web → qa-testing → security-baseline → launch-runbook
↓
domain-strategy → monitoring-and-alerting → backup-and-disaster-recovery
↓
incident-response → after-action-report
↓
analytics-strategy → cro-optimization → ux-research → usability-testing → journey-mappingThe SEO audit suite (Ahrefs MCP-powered) wraps around the SEO foundation skills:
seo-audit-orchestration
├── seo-site-health-audit
├── seo-backlink-audit
├── seo-keyword-gap-audit
├── seo-content-gap-audit
├── seo-traffic-diagnosis (also runs standalone for incident-style work)
└── seo-rank-tracking (ongoing, feeds the others)The catalog also includes four audience tracks that compose alongside the foundational lifecycle. Each track has its own internal flow:
Paid media (Marketing track):
paid-media-strategy → ads-creative-development → ads-performance-analyticsPairs with the paid media platforms in the integrations catalog at rampstack.co (Google Ads, Meta, LinkedIn, TikTok, plus Synter as the multi-platform aggregator).
Growth tooling (interactive web tools):
funnel-flow-architecture (orchestrator)
├── lead-magnet-design (capture)
├── calculator-design (capture / activate)
├── quiz-and-assessment-design (capture / activate)
├── multi-step-form-design (activate)
├── chatbot-flow-design (activate)
├── onboarding-wizard-design (activate)
├── interactive-product-tour (activate / convert)
├── upgrade-flow-design (convert)
├── scheduler-and-booking-design (convert)
├── comparison-tool-design (convert)
└── product-configurator-design (convert)funnel-flow-architecture is the orchestrator: it sequences which interactive tool fits each audience and funnel stage, distinguishing matched-funnels from kitchen-sink-funnels.
Tier 2 content lifecycle:
content-strategy → pillar-content-architecture → content-brief-authoring
↓
content-and-copy / long-form-content-frameworks / email-sequences
↓
editorial-qa → content-distribution → programmatic-seo
↓
content-refresh-system → content-repurposing → content-migrationai-content-collaboration is a workflow layer that runs across every phase rather than a single step. documentation-strategy operates continuously alongside the rest.
Tier 2 product management (two parallel tracks):
Experimentation track:
experiment-design → feature-flagging → experimentation-platform-orchestrator
↓
experimentation-analytics → data-warehouse-experimentation
Gap-closing track:
pm-spec-writing → roadmap-planning → feature-launch-playbook
↓
beta-program-management → product-analytics-setup → integration-orchestratorThe experimentation track ships changes with statistical discipline; the gap-closing track ships features with operational discipline. Both compose with the foundational lifecycle above.
Operations, cross-cutting, and team skills (stakeholder-communication, documentation-strategy, vendor-evaluation, team-onboarding-playbook, dependency-management, cost-optimization, etc.) cut across every track.
You can also pull individual skills for one-off work. Need just a backlink audit? Use seo-backlink-audit. Need to write a creative brief? Use creative-brief. Each skill stands on its own.
How the catalog connects
The skills compose with the tools your team already uses. The catalog sits at the center; 35 integrations across 6 integration categories radiating out via MCPs.
<p align="center"> <picture> <source media="(max-width: 640px)" srcset="docs/architecture-mobile.jpg"> <img src="docs/architecture-wide.jpg" alt="RampStack architecture diagram. A central navy hub card shows the RampStack mark with the subtitle 'Stack-agnostic methodology'. Six category cards radiate out: Workflow with 6 integrations (Jira, Linear, Notion, Figma, GitHub), Experimentation with 11 integrations (Statsig, PostHog, Optimizely, Amplitude), SEO Intelligence with 3 integrations (Ahrefs, Semrush, Similarweb), Paid Media with 5 integrations (Google Ads, Meta Ads, LinkedIn, TikTok, Synter), Content and SEO with 5 integrations (Webflow, Contentful, Frase, Profound, AirOps), and Data and Analytics with 5 integrations (BigQuery, Snowflake, Mixpanel, dbt, Hex)."> </picture> </p>
Surfaces
<!-- SURFACES:START --> This catalog is the open-source methodology layer. Commercial surfaces at rampstack.co extend it:
- [Skills directory](https://rampstack.co/skills). Every skill on a curated landing surface with audience tracks, search, and category navigation.
- [Walkthroughs](https://rampstack.co/walkthroughs). Multi-skill recipes that orchestrate skill clusters end-to-end. Use these when one skill is not enough and a packaged sequence is.
- [Integrations directory](https://rampstack.co/integrations). Curated MCPs, APIs, and tooling that the skills hook into.
- [Showcase](https://rampstack.co/showcase). Real brand sites built from these skills, with the brief that produced each one.
The skills in this repository remain free, open-source, and stack-agnostic. The surfaces above are how the same methodology is delivered as a product. <!-- SURFACES:END -->
Design principles
claude-skills follows the Agent Skills Specification, the open standard for portable AI agent skills originally developed by Anthropic and adopted across the AI tooling ecosystem (Claude Code, OpenAI Codex, Gemini CLI, GitHub Copilot, Cursor, VS Code, Goose, Spring AI, and 30+ other platforms as of early 2026).
Beyond the format itself, the catalog is designed around three principles aligned with the guidance Anthropic publishes in Building effective agents:
Simplicity. Each skill covers one focused capability rather than trying to be a multi-purpose document. A roadmap-planning skill plans roadmaps. A keyword-research skill researches keywords. Composing them together produces complex workflows; mixing them inside one skill produces unreliable ones.
Transparency. Every skill declares its scope, dependencies, and expected behavior in machine-readable YAML frontmatter. The catalog is inspectable by tooling, not just by humans reading prose.
Quality contracts via tooling. Structural and content quality is enforced through automated checks (run python .github/scripts/lint_skills.py) rather than convention alone. Every skill is validated against a schema. Every catalog change is validated in CI.
Skills in this catalog are designed to compose into the common agentic workflow patterns Anthropic documents: prompt chaining (sequential steps), routing (classify and direct), parallelization (sectioning or voting), orchestrator-workers (dynamic delegation), and evaluator-optimizer (iterative refinement).
Because the catalog conforms to the open Agent Skills standard, skills work across any platform supporting the specification without modification.
Family repos
claude-skills is the parent catalog. Curated subsets and companion repos focus on specific specialties:
| Repo | Focus | |
|---|---|---|
| claude-skills | Full catalog (you are here) | |
| claude-skills-starter | General-purpose lite | |
| claude-skills-seo | SEO consulting | |
| claude-skills-pm | Product management | |
| claude-skills-widgets | UI patterns + components | |
| awesome-claude-skills | Curated discovery list | n/a |
Each family repo is MIT-licensed, conforms to the Agent Skills Specification, and is stack-agnostic. Use the full catalog for breadth; use a specialty subset when working in one domain.
How this org fits together
The table above covers the skill catalogs. They are one part of a larger set, and the rest of it is below. All of it is public.
Skills. This repo is the canonical home for all skill content. Alongside it sits the workflows tier: fifteen multi-skill runbooks with their connectors, a getting-started guide, and published run records for the ones that have been executed as written.
Subsets. The five curated repos in the table above copy from this catalog with attribution and track it upstream.
Design direction themes. Thirteen sibling repos, each shipping annotated design tokens with their measured contrast ratios, a component layer, two Tailwind adapters, and a demo that opens from a file with nothing installed. They come in three artifact classes: seven surface registers, one layout archetype, and five shells, which ship a structure a site lives inside (a window manager or a board, a taskbar or a dock, an enhancement contract and a focus model) with the register they wear left swappable. VivaOcean, the animated-scene shell, is the showcase flagship. What the shell class settled, and why, is public in its class decision log, which every new shell reads first and continues. All thirteen are linked from the gallery at rampstack.co/themes.
Creative direction. The themes are not thirteen moods. Each one states its coordinates in the creative direction framework, which sets brand direction on four axes, and the showcase renders archetypes at each position on it.
Engines. Krine, Tholo, and Basano run on one runtime: Krine decides, Tholo builds, Basano proves. The engines page covers what the three share.
Research. The SERP event registry is a dated, sourced, confidence-tagged record of AI model releases, search feature changes, and confirmed algorithm updates, rendered on the site from the repository that holds it.
What shipped, and when, is recorded at rampstack.co/updates.
<!-- COUNT_CATALOG_HEADER:START -->
The 107-skill catalog
<!-- COUNT_CATALOG_HEADER:END -->
<!-- COUNT_CATALOG_INTRO:START --> All 107 skills are shipped. Each has a complete SKILL.md plus at least one reference file (template, checklist, or playbook). <!-- COUNT_CATALOG_INTRO:END -->
<!-- AUTO-GENERATED CATALOG: do not edit by hand. Run scripts/generate_readme_catalog.py --write --> <!-- CATALOG:START -->
Strategy and discovery (5)
| Skill | What it does |
|---|---|
brand-discovery | Audience research, competitive scan, positioning territory exploration<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
creative-brief | Project briefs that align stakeholders before work starts<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
creative-direction | Four-axis aesthetic brief (tone, aesthetic, audience, sensory ambition) for cross-skill coherence<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
information-architecture | Sitemap, navigation, URL structure, content types, taxonomy<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
content-strategy | Editorial strategy, content calendar, topical authority planning<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
Brand (7)
| Skill | What it does |
|---|---|
brand-ideation | Naming, positioning territories, mood directions, narrative angles<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
brand-identity | Logo system, color, typography, imagery, iconography, motion<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
brand-style-guide | The canonical reference document for the full brand system<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
brand-voice | Voice attributes, tone shifts, vocabulary, paired-example library<br>Tested on OpenAddict (newest run 2026-08-30). Via API: Claude Haiku 4.5 (No measured effect), Gemini 3.1 Flash Lite (Holds), GPT-5 mini (No measured effect). Per model. |
brand-archetype-system | 12 archetype defaults across 18 verticals: color, type, voice, imagery starters<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
logo-design | Logo variants across architectures (wordmark, lockup, monogram, letterform-as-symbol), with rationale and application specs<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
creative-brief-selector | Live-reference-grounded creative briefs with divergence check against prior builds<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
Design (4)
| Skill | What it does |
|---|---|
design-system | Component library, design tokens, design system documentation<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
design-standards | Production-grade page and component design standards<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
art-direction | Photography, illustration, and visual direction for campaigns<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
vertical-site-conventions | Vertical page and site composition built to the experience bar<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
Content (13)
| Skill | What it does |
|---|---|
pillar-content-architecture | Hub-level content architecture: pillar topic selection, cluster planning, internal linking, URL structure, pillar and cluster page anatomy, topical authority signals, refresh discipline<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
content-brief-authoring | Per-piece editorial brief: target keyword, intent, audience, outline, entity coverage, internal linking, success criteria, and the discipline that distinguishes useful briefs from bloat<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
content-and-copy | Website copy, blog content, content production frameworks<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
landing-page-copy | Landing pages, sales pages, hero-to-CTA flow<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
email-sequences | Onboarding flows, lifecycle campaigns, transactional copy<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
programmatic-seo | Designing pSEO programs that work: data sources, template design, quality control at scale, internal linking, crawl budget, AEO/GEO patterns, refresh discipline, and when pSEO is and is not the right answer<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
editorial-qa | Pre-publish QA framework: brief adherence, voice consistency, fact accuracy, AI-content audit, AEO/SEO compliance, sampling at scale, and the workflow that distinguishes catch-problems QA from process theater<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
ai-content-collaboration | How humans and AI compose in content workflows: participation boundaries, hybrid patterns, voice ownership, the AI slop problem, disclosure and transparency, team calibration, and the ethics of honest AI-assisted production<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
long-form-content-frameworks | Structural patterns for individual long-form pieces (case studies, whitepapers, research reports, definitive guides, manifestos, ebooks, long-form tutorials) that distinguish publication-quality work from bloggy-long padding or academic bloat<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
content-refresh-system | Systematic content refresh: quarterly audits, refresh prioritization, refresh-vs-merge-vs-delete decisions, the lifecycle discipline that distinguishes intentional programs from set-and-forget decay<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
content-repurposing | Cross-format content adaptation: one piece becomes many (blog series, email, social, webinar, podcast, video) with per-format adaptation rather than mass-blast that ignores medium constraints<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
content-distribution | Content distribution discipline: owned, earned, and paid channels matched to audience and content type. Channel-fit decisions, distribution cadence, the strategic alternative to spam-everywhere or hope-and-pray<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
evidence-based-reviews | Evidence tiers, methodology disclosure, honest review claims<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
SEO foundation (7)
Tool-agnostic SEO skills. These define the conceptual frameworks. The SEO audit suite below adds the Ahrefs MCP-powered execution layer.
| Skill | What it does |
|---|---|
seo-onpage | Single-page audits and optimization across 8 dimensions<br>Tested on OpenAddict (Tier 1, newest run 2026-09-26). Via API: Claude Haiku 4.5 (No measured effect), Gemini 3.1 Flash Lite (Mixed), GPT-5 mini (No measured effect). In Claude Code: Claude Haiku 4.5 (No measured effect), Claude Fable 5 (No measured effect), Claude Opus 5 (No measured effect), Claude Sonnet 5 (No measured effect), Claude Fable 5.1 (No measured effect), Claude Opus 5.5 (No measured effect). Per model. |
seo-technical | Crawlability, indexability, rendering, schema, page experience<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
seo-keyword | Discovery, intent classification, clustering, prioritization<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
seo-competitor | SERP overlap, content gaps, backlink gaps, technical comparison<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
seo-offpage | Link building, digital PR, citations, linkable assets<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
seo-content-audit | Keep/update/merge/redirect/delete decisions across a site<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
seo-aeo-geo | AI search optimization, llms.txt, extraction-friendly content<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
SEO audit suite (Ahrefs MCP-powered) (7)
End-to-end SEO audit workflows that pull data from the Ahrefs MCP and produce concrete deliverables. These skills assume the Ahrefs MCP is connected.
| Skill | What it does |
|---|---|
seo-audit-orchestration | Master orchestrator: sequences the suite, produces a rollup report<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
seo-backlink-audit | Profile health, anchor mix, toxic links, reclamation, gap analysis<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
seo-keyword-gap-audit | Competitor keyword gaps with opportunity scoring and clustering<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
seo-content-gap-audit | Missing topics, thin coverage, outdated content, decay diagnosis<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
seo-traffic-diagnosis | Diagnose drops, stalls, or wins via 5-layer root cause analysis<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
seo-site-health-audit | Triage Ahrefs Site Audit findings by SEO impact, not severity<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
seo-rank-tracking | Setup, baseline, segmentation, alerting, dashboarding<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
Product (14)
| Skill | What it does |
|---|---|
pm-spec-writing | PRDs, user stories, acceptance criteria, dev briefs<br>Tested on OpenAddict (Tier 1, newest run 2026-09-26). Via API: Claude Haiku 4.5 (Mixed), Gemini 3.1 Flash Lite (Mixed), GPT-5 mini (Mixed). In Claude Code: Claude Haiku 4.5 (No measured effect), Claude Fable 5 (No measured effect), Claude Opus 5 (No measured effect), Claude Sonnet 5 (No measured effect), Claude Fable 5.1 (No measured effect), Claude Opus 5.5 (Could not measure). Per model. |
ticket-authoring | Bugs, tasks, stories, spikes, and epics with a definition of ready per type<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
roadmap-planning | Quarterly planning, prioritization, dependency mapping<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
integration-orchestrator | Sequence creative-direction work across phases, gates, handoffs, and QA verification<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
experiment-design | Hypothesis to decision: sample size, duration, segment analysis, interpretation, and the failure modes that produce wrong shipping calls<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
feature-flagging | Flags as production infrastructure: types, naming, lifecycle, targeting, rollout, stale flag cleanup, governance<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
experimentation-analytics | Read result panels without fooling yourself: confidence intervals, p-values, multiple testing, sequential testing, CUPED, ratio metrics, network effects, dashboard reconciliation<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
experimentation-platform-orchestrator | Pick the right experimentation platform, migrate when wrong, coordinate when multi-platform: a decision framework for Statsig, PostHog, GrowthBook, Optimizely, Amplitude, Eppo, Kameleoon<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
product-analytics-setup | Instrument product analytics correctly: event taxonomy, properties, naming conventions, schema versioning, funnels, retention cohorts, North Star selection, and the instrumentation debt that compounds without discipline<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
data-warehouse-experimentation | Run experiments out of the warehouse: SQL assignment, exposure logs, dbt metric definitions, statistical analysis, variance reduction with CUPED, sequential testing, and the operational tradeoffs vs platforms<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
feature-launch-playbook | The operational discipline of launching a feature well: positioning, internal alignment, customer comms, sales enablement, support readiness, rollout strategy, monitoring, and post-launch measurement<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
jtbd-framing | Jobs-to-be-Done framework. Job statements, struggling moments, hire/fire criteria, the difference between feature-thinking and job-thinking. Honest about where JTBD earns its keep and where it becomes performative<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
okr-design | OKR design discipline. Outcome statements, key results, scoring, mid-quarter recalibration. Distinguishes sandbagged OKRs (always hit, useless) from aspirational fantasy (impossible, demoralizing) from stretch OKRs (genuine ambition with quarterly accountability)<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
beta-program-management | Running betas that produce real signal. Participant selection, structured feedback, beta-to-GA decisions. Distinguishes soft-launch (no structure) from kitchen-sink (everyone in) from structured-beta (calibrated cohort with intentional feedback loops)<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
Development (5)
| Skill | What it does |
|---|---|
code-review-web | PR review, build error diagnosis, security and quality checks<br>Tested on OpenAddict (Tier 1, newest run 2026-09-26). Via API: Gemini 3.1 Flash Lite (No measured effect), GPT-5 mini (No measured effect). In Claude Code: Claude Fable 5.1 (No measured effect), Claude Haiku 4.5 (No measured effect), Claude Opus 5 (No measured effect), Claude Sonnet 5 (No measured effect), Claude Fable 5 (No measured effect), Claude Opus 5.5 (No measured effect). Per model. |
frontend-component-build | Component architecture, props design, accessibility from the start<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
accessibility-audit | WCAG compliance audit with remediation plan<br>Tested on OpenAddict (Tier 1, newest run 2026-09-26). Via API: Claude Haiku 4.5 (Holds), Gemini 3.1 Flash Lite (No measured effect), GPT-5 mini (No measured effect). In Claude Code: Claude Haiku 4.5 (Mixed), Claude Opus 5 (No measured effect), Claude Fable 5 (Mixed), Claude Sonnet 5 (No measured effect), Claude Fable 5.1 (No measured effect), Claude Opus 5.5 (No measured effect). Per model. |
performance-optimization | Core Web Vitals, asset optimization, render performance<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
oss-contribution-research | Gated research and patch prep for outside open-source PRs, ending in a human-approved queue item<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
Quality assurance (1)
| Skill | What it does |
|---|---|
qa-testing | Pre-launch QA, regression testing, cross-browser checks<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
Operations (9)
| Skill | What it does |
|---|---|
launch-runbook | Go-live runbook, DNS cutover, deploy day procedures<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
incident-response | Incident triage, comms, mitigation, escalation<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
after-action-report | Post-mortems, retros, learnings documentation<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
domain-strategy | DNS architecture, redirects, registrars, multi-domain portfolios<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
monitoring-and-alerting | SLO design, uptime checks, alert routing, on-call rotations<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
backup-and-disaster-recovery | RPO/RTO targets, backup strategy, restoration drills<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
security-baseline | HTTPS, security headers, CSP, secrets management, vulnerability scans<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
email-deliverability | DMARC, SPF, DKIM, sender reputation, deliverability monitoring<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
media-asset-management | Image pipelines, video hosting, asset libraries, format selection<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
Growth (2)
| Skill | What it does |
|---|---|
analytics-strategy | Measurement frameworks, dashboard design, event taxonomy<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
cro-optimization | Hypothesis-driven testing, conversion optimization<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
Growth tooling (12)
Interactive web tools that turn visitors into leads. Lead magnets, calculators, quizzes, multi-step forms, chatbots, and the cross-tool funnel architecture that orchestrates them.
| Skill | What it does |
|---|---|
lead-magnet-design | Designing gated content that earns the email. Distinguishes thin-bait (overpromises, underdelivers) from kitchen-sink-resource (everything, helps with nothing) from earned-value-magnet (delivers standalone value while qualifying the lead)<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
calculator-design | Designing interactive calculators that deliver decision-support value while qualifying leads. Distinguishes vanity-calculator (no real value) from lead-trap (hides answer behind email) from transparent-decision-tool (gives genuine value, captures leads honestly)<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
quiz-and-assessment-design | Designing quizzes and assessments that produce actionable segmentation. Distinguishes clickbait-quiz (engagement only) from vanity-result (entertaining, not useful) from actionable-segmentation (genuine categorization that drives next-step recommendations)<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
multi-step-form-design | Designing multi-step forms that respect cognitive load while maintaining completion intent. Distinguishes kitchen-sink-single-page (overwhelms) from progress-theater (steps without genuine staging) from genuinely-staged (each step earns its own page)<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
chatbot-flow-design | Designing conversational flows for chatbots and AI agents on websites. Distinguishes scripted-bot (rigid trees, fail edge cases) from hallucinating-bot (LLM without structure, makes things up) from structured-guided-conversation (LLM-powered with intent architecture and fallback discipline)<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
funnel-flow-architecture | Architecting cross-tool conversion flows that match audience and stage. Distinguishes silo-funnels (every tool standalone) from kitchen-sink-funnels (every audience squeezed through one path) from matched-funnels (architecture matched to audience-and-stage)<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
onboarding-wizard-design | Designing first-run product onboarding wizards. Distinguishes tutorial-overload (dump everything upfront) from skip-friendly-empty (skipped onboarding leads to abandoned product) from earned-progressive-disclosure (right things at the right moments)<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
interactive-product-tour | Designing in-product tours and contextual help. Distinguishes tooltip-spam (every button has a tour stop) from one-and-done (tour shows once, never seen again) from contextual-when-needed (surfaces help at the moment friction occurs)<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
upgrade-flow-design | Designing free-to-paid conversion flows. Distinguishes paywall-everywhere (gates everything aggressively) from free-forever-trap (no upgrade path surfaces) from value-triggered-upgrade (paywall surfaces at moments of demonstrated value)<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
scheduler-and-booking-design | Designing schedulers and booking flows. Distinguishes any-time-friction (no qualification, just a booking link) from interrogation-gate (so much qualification it scares users off) from qualified-fast-path (just enough qualification to set up the call well)<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
comparison-tool-design | Designing comparison tools that help users decide. Distinguishes feature-list-dump (every feature in a row, no decision support) from hidden-recommendation (biased comparison pretending to be neutral) from honest-comparison-with-guidance (genuine comparison plus opinionated recommendation)<br>Tested on OpenAddict: set up for testing, not yet run (coverage). |
product-configurator-design | Designing interactive product configurators. Distinguishes infinite-options (decision paralysis from too many options) from canned-bundles-only (no real customization) from guided-configuration (smart defaults plus meaningful const |