Design and run statistically valid A/B tests and build a growth experimentation program: hypotheses, sample sizes, variant design, analysis, ICE-prioritized backlogs. Use when the user mentions split tests, experiments, statistical significance, "which version is better," or experiment velocity. Not for generating page-level optimization ideas: use cro. Not for tracking implementation: use analytics. Not for writing variant copy: use copywriting.
Design a reusable multi-skill recipe, a documented loop blueprint that sits between a one-off skill and a scheduled automation. Use when turning a workflow (lead enrichment, newsletter issue, SEO refresh, proposal generation) into a repeatable agent loop, or before promoting any workflow to a schedule. Not for auditing existing automations: use automation-audit-ops. Not for verifying a recipe's output: use behavior-validator.
Self-rate completed work on five axes (accuracy, completeness, clarity, actionability, conciseness) with cited evidence per score. Use after multi-file code changes, multi-step workflows, long debugging sessions, or written analysis, or when asked "rate yourself". Not for proving external behavior works: use behavior-validator. Not for comparing agents head to head.
Contracts for integrating paid AI/ML providers behind swappable adapters: typed interfaces, idempotent job lifecycle, cost reservation, retry classes, health monitoring. Use when adding a model vendor, wiring paid generation APIs, handling provider webhooks, or tracking per-generation spend. Not for writing prompts that direct AI coding tools: use ai-tool-directive.
Optimize content to be cited by AI search engines and assistants (AI Overviews, ChatGPT, Perplexity, Claude, Gemini, Copilot). Use when the user says AI SEO, AEO, GEO, LLMO, answer engine optimization, AI citations, AI visibility, llms.txt, or "show up in AI answers". Not for traditional technical or on-page audits: use seo-audit. Not for structured data implementation: use schema. Not for building the content plan itself: use content-strategy.
Write structured directives for AI coding tools so fixes remove old code, respect scope, and verify. Use when delegating technical work to another AI assistant, especially when past attempts layered new code on old, missed scope, or failed to stick. Not for integrating AI provider APIs into a product: use ai-provider-integration.
Set up, improve, or audit analytics tracking and measurement. Use when the user mentions GA4, Google Analytics, GTM, tag manager, conversion tracking, event tracking, UTM parameters, tracking plan, Mixpanel, Segment, attribution, "are my events firing", or "how do I measure this". Not for experiment design or A/B test measurement: use ab-testing. Not for organic traffic analysis: use seo-audit. Not for conversion optimization work that consumes this data: use cro.
Design stable interfaces that are hard to misuse. Use when creating REST or GraphQL endpoints, defining type contracts between modules, designing component props, or changing any public interface. Not for retiring or breaking an existing contract: use deprecation-and-migration. Not for provider adapter patterns around paid AI APIs: use ai-provider-integration.
Evidence-first inventory and overlap audit of automations, scheduled jobs, CI workflows, connectors, and agent loops. Use when the user asks what automations exist, what is live, what is broken, or what overlaps, before fixing anything. Not for designing new loops: use agent-recipe-builder. Not for proving a single workflow's output works: use behavior-validator.
Build editorial, award-tier public pages: opinionated type, asymmetric layout, one motion moment per viewport, ruthless copy reduction on a web-standards floor. Use when creating or restyling landing, marketing, portfolio, pricing, about, or blog index pages, or when asked for premium, editorial, studio-quality design. Not for internal app UI (admin, dashboard, settings, console, API routes). Not for measuring decay in an existing UI: use design-system-drift.
Subscribe contacts to a Beehiiv newsletter publication via the Beehiiv v2 API, from forms, scripts, or automations. Use when pushing email addresses into Beehiiv, adding subscribers programmatically, or wiring a signup form to a publication. Not for publishing to Ghost, Substack, or Listmonk: use ghost-publisher, substack-publisher, or listmonk-publisher. Not for writing the newsletter content: use copywriting or emails.
Verify apps, CLIs, API routes, artifacts, and automations from the user-facing surface, not from implementation claims. Use when checking whether something actually works, catching false-success states, missing side effects, or workflow regressions. Not for visual and layout QA in a real browser: use browser-visual-qa. Not for rating output quality: use agent-self-evaluation.
Turn a raw idea into an approved design before any code exists. Use before creating features, building components, adding functionality, or modifying behavior, whenever intent and requirements are not yet pinned down. Not for writing the implementation plan: use writing-plans after the design is approved. Not for validating finished work: use behavior-validator.
Autonomous visual and interaction QA on real UI in a real browser: drive the page like a user, capture screenshots and accessibility snapshots, diff against baselines, read console errors and Core Web Vitals. Use to verify frontend work before calling it done, or to catch layout breakage, dead states, and regressions. Not for backend or non-visual behavior checks: use behavior-validator. Not for CI pipeline setup: use ci-cd-and-automation.
Architecture for multi-mechanic, persistent 3D browser games in React Three Fiber: system/view separation, fixed-timestep loop, AI state machines, instancing, cloud saves. Use when building open-world exploration, gathering loops, mounts, enemy AI, inventory, or leaderboards. Not for marketing 3D heroes: use build-3d-website. Not for world polish or character/horse systems: use high-end-browser-western and high-end-character-and-horse.
Build interactive 3D marketing sites with React Three Fiber and drei: 3D heroes, floating objects, tilt cards, scroll-driven scenes, WebGL showcases. Use when asked for a 3D landing page, WebGL hero, immersive explorer, or a premium wow page, or on mentions of three.js, R3F, tilt card. Not for playable games with mechanics, AI, or saves: use build-3d-game.
Drive a captured goal to 100% completion through a bounded implement-verify loop with hard caps and halt-on-ambiguity. Use when the user says loop until done, keep building, autonomous build, drive to completion, don't stop until it's done. Requires .lovable/goal.json. Not for defining the goal or acceptance criteria: run goal-capture first.
Set up and enforce CI/CD pipelines: quality gates, test runners in CI, deployment strategies, rollback. Use when creating or modifying build and deploy pipelines, adding automated checks, or debugging CI failures. Not for writing the tests themselves: use test-driven-development. Not for diagnosing an app bug CI surfaced: use systematic-debugging. Not for branch and merge mechanics: use git-workflow-and-versioning.
SEO analysis orchestrator, routes audits to the right specialist skill and merges results into one scored report. Use when the user says "SEO audit", "analyze my site's SEO", "SEO health", or asks a broad SEO question with no obvious specialist. Not for narrow jobs you can route directly: schema goes to claude-seo-schema, sitemaps to claude-seo-sitemap, crawl/CWV to claude-seo-technical, intent fit to claude-seo-sxo.
Backlink profile analysis: referring domains, anchor text distribution, toxic link detection, competitor link gap. Use when the user says backlinks, link profile, referring domains, anchor text, toxic links, link gap, link building, disavow, or backlink audit. Not for keyword clustering or content architecture: use claude-seo-cluster. Not for building the overall SEO strategy: use claude-seo-plan.
SERP-overlap topic clustering for content architecture: group keywords by shared Google top-10 results, design hub-and-spoke clusters with internal link matrices, output briefs and a cluster map. Use when the user says topic cluster, content cluster, pillar page, hub and spoke, keyword grouping, or content architecture. Not for backlink analysis: use claude-seo-backlinks. Not for the overall SEO strategy document: use claude-seo-plan. Not for optimizing a single page: use claude-seo-page.
Build SEO comparison and alternatives pages ("X vs Y", "alternatives to X", category roundups) with feature matrices, schema, and honest competitor data. Use when the user says "comparison page", "vs page", "alternatives page", "X vs Y", or "alternative to". Not for full content briefs on other page types (use claude-seo-content-brief) or generating hundreds of comparison pages from data (use claude-seo-programmatic).
Generate competitive SEO content briefs with competitor gap scoring, per-section word counts, keyword placement, and meta tags. Handles new-page briefs and improve-existing-page briefs. Use when the user says "content brief", "content outline", "blog brief", "writing brief", or "outline for". Not for site-wide strategy or content calendars (use claude-seo-plan) or comparison pages specifically (use claude-seo-competitor-pages).
Pull live SEO data through the DataForSEO MCP server, including SERPs, keyword metrics, backlinks, on-page audits, competitor domains, business listings, and LLM mention tracking. Use when the user says "dataforseo", "live SERP", "keyword volume", "backlink data", "competitor data", or "AI visibility check". Not for Google first-party data like Search Console, CrUX, or GA4 (use claude-seo-google) or maps-platform analysis (use claude-seo-maps).
Capture baselines of SEO-critical page elements, diff current state against them, and classify regressions by severity. Git for on-page SEO. Use when the user says "SEO drift", "baseline", "did anything break", "SEO regression", "before and after", or "deployment check". Not for one-off page audits with no baseline (use claude-seo-page) or full technical crawls (use claude-seo-technical).
Run the FLOW operating loop for evidence-led SEO, Find then Leverage then Optimize then Win, routing each stage to the right analysis and keeping recommendations tied to evidence. Use when the user says "FLOW", "FLOW framework", "evidence-led SEO", or asks where to start an SEO engagement. Not for executing a single specialized audit directly (go straight to claude-seo-page, claude-seo-geo, claude-seo-local, or claude-seo-technical).
Audit and optimize content for AI search surfaces, including Google AI Overviews and AI Mode, ChatGPT search, and Perplexity, covering citability, structure, crawler access, and brand mention signals. Use when the user says "GEO", "AI Overviews", "AI search", "LLM optimization", "AI citations", or "AI visibility". Not for local AI visibility specifics (use claude-seo-local) or pulling live LLM mention data (use claude-seo-dataforseo).
Pull Google first-party SEO data via Search Console, PageSpeed Insights, CrUX, Indexing API, GA4, YouTube Data, Knowledge Graph, and Cloud NLP. Real field data for Core Web Vitals, indexation, search performance, and organic traffic. Use when the user says "Search Console", "GSC", "PageSpeed", "CrUX", "field data", "URL inspection", "GA4 organic", "LCP", "INP", or "CLS". Not for third-party SERP, keyword, or backlink data (use claude-seo-dataforseo).
Audit local SEO signals on a business website, including GBP integration, reviews, NAP consistency, citations, LocalBusiness schema, location pages, and local link signals, with business-type and industry detection. Use when the user says "local SEO", "GBP", "map pack", "citations", "NAP", "service area", or "multi-location". Not for maps-platform data like geo-grids and GBP API audits (use claude-seo-maps) or general AI visibility (use claude-seo-geo).
Analyze a business's presence on maps platforms, including geo-grid rank tracking, GBP profile audits via API, cross-platform review intelligence, NAP verification across Google/Bing/Apple/OSM, competitor radius mapping, and schema generation. Use when the user says "geo-grid", "maps analysis", "GBP audit", "review velocity", "competitor radius", "Share of Local Voice", or "SoLV". Not for on-website local SEO signals (use claude-seo-local) or AI search visibility (use claude-seo-geo).
Deep SEO analysis of a single page, covering on-page elements, content quality, technical meta tags, schema, images, and CWV risk flags, with a scored report. Use when the user says "analyze this page", "check page SEO", "page analysis", or provides one URL for review. Not for site-wide crawls or indexability audits (use claude-seo-technical), regression tracking against a baseline (use claude-seo-drift), or AI-surface scoring (use claude-seo-geo).
Build a strategic SEO plan for a new or existing site, covering discovery, competitive analysis, site architecture, content strategy, and a phased implementation roadmap with KPIs. Use when the user says "SEO plan", "SEO strategy", "SEO roadmap", "content calendar", or "site architecture". Not for a single page brief (use claude-seo-content-brief), pages generated at scale (use claude-seo-programmatic), or executing audits on a live site (use claude-seo-page or claude-seo-technical).
Plan and audit SEO pages generated at scale from data sources, enforcing quality gates against thin content, doorway patterns, and index bloat. Use when the user says "programmatic SEO", "pages at scale", "template pages", "generated pages", or "data-driven pages". Not for one-off page audits (use claude-seo-page), site strategy (use claude-seo-plan), or hand-built comparison pages (use claude-seo-competitor-pages).
Detect, validate, and generate Schema.org structured data as JSON-LD. Use when the user says "schema", "structured data", "rich results", "JSON-LD", or "markup". Not for full technical audits: use claude-seo-technical. Not for site-wide SEO scoring or orchestration: use claude-seo.
Analyze existing XML sitemaps or generate new ones, validating format, URLs, and structure. Use when the user says "sitemap", "generate sitemap", "sitemap issues", or "XML sitemap". Not for full crawl or indexability audits: use claude-seo-technical. Not for structured data: use claude-seo-schema.
Search Experience Optimization, reads the SERP backwards to find page-type mismatches, derive user stories, and score pages per persona. Use when the user says "SXO", "why isn't my page ranking", "intent mismatch", "SERP analysis", "persona scoring", or "page type mismatch". Not for technical health checks: use claude-seo-technical. Not for full-site audits: use claude-seo.