Production-ready Claude Agent Skills for SEO, powered by the SE Ranking remote MCP. Each skill turns raw API data into a finished SEO deliverable — content briefs, AI Search share of voice, page intelligence, structured data, drift monitoring, SXO diagnostics, competitive analysis, and more.

These Claude Skills are designed to work with the SE Ranking MCP server, but they document every API call explicitly so they can also be adapted to other SEO data providers.

Three ways. Pick the first one that fits.

The easiest setup — no terminal needed. Cowork in Claude Desktop installs the plugin from a marketplace through a UI flow.

1. Open Customize in the sidebar.
2. Click Personal plugin → + Create plugin.
3. Click Add marketplace and enter seranking/seo-skills.
4. Install the plugin once the marketplace loads.

Skills are available in your next Cowork session. Trigger them by name — seo-content-brief, seo-page, etc. — or just describe what you want and Claude picks the right one.

If you use Claude Code in the terminal:

/plugin marketplace add seranking/seo-skills
/plugin install seo-skills@seranking

Trigger skills with /seo-content-brief, /seo-page, etc. — no seo-skills: prefix needed. Or just describe what you want and Claude picks the right skill. Update the marketplace later with /plugin marketplace update seranking.

Options 1 and 2 cover 25 of 26 skills. The 26th — seo-google — ships Python scripts that need a local clone. The optional Firecrawl + Google extensions also need this. If neither matters to you, skip this option.

git clone --depth 1 https://github.com/seranking/seo-skills.git
bash seo-skills/install.sh

The installer is interactive — it asks which extensions to install. Accept the defaults and you're done. Re-running pulls the latest version and re-runs the chosen extensions. Requires git, python3 3.10+, and (for Firecrawl) node 20+.

curl -fsSL https://raw.githubusercontent.com/seranking/seo-skills/main/install.sh | bash -s -- --all

The manual install only wires your environment (extensions, Python deps, ~/.config/seo-skills/). You still need Option 1 or 2 above so Claude can find the skill files.

The SE Ranking remote MCP auto-registers when you install the plugin — there's no separate claude mcp add step. On your first session, run /mcp and sign in via OAuth. No API token to manage. If you don't have an SE Ranking account yet, sign up — API access is required.

claude mcp add --transport http se-ranking https://api.seranking.com/mcp

25 finished example outputs — covering 25 of the 26 skills — live in examples/. Each is a real run against a publicly observable target. Three flagship examples to skim before installing:

• seo-plan for wix.com (90-day plan) — phased SEO roadmap that composes outputs from seo-ai-search-share-of-voice, seo-competitor-gap-analysis, and seo-drift into a 90-day plan with owners, success metrics, and a critical path. Demonstrates skill composition.
• seo-technical-audit for linear.app — top-10 priority fix list ranked by severity × affected_pages / effort, plus issues.csv for the engineering team and a per-key-page issue digest.
• seo-ai-search-share-of-voice for wix.com — 36.1% combined share of voice across 5 LLM engines (AI Overview / ChatGPT / Perplexity / Gemini / AI Mode), heatmap, topic-cluster ownership grid, and 5 prioritised gap-closing actions.

Full index of all 25 examples: examples/README.md. Every example is dated and snapshot-disclaimed — re-run the skill for fresh numbers.

The seven to reach for first — each links to a real, dated example run.

The rest of the catalogue — every other skill the plugin ships.

Some skills work best when paired with optional MCP servers or Python toolchains beyond SE Ranking. Each extension is opt-in — the skills that use it degrade gracefully when it's absent: they fall back to a lower-fidelity method, or note what was unavailable, rather than failing the run.

The fastest way to install both extensions is the top-level install.sh (Install option 3). The per-extension scripts below are equivalent — use them when you only want one of the two, or when you've already cloned the repo and want to install/re-install a single extension without re-running the wrapper.

Firecrawl is the recommended way to run any skill that inspects raw HTML — but it is not mandatory. Without it, those skills fall back to WebFetch (Claude's built-in page fetch) and still produce a deliverable, just a lower-fidelity one: WebFetch returns markdown only, so every <head> <meta> tag, every <link rel="canonical">, and every <script type="application/ld+json"> block is stripped before the skill sees it. The Firecrawl extension closes that gap and adds full-site crawl, URL discovery, in-site search, and screenshots.

# From inside a clone of this repo:
bash extensions/firecrawl/install.sh

The script verifies Node 20+, prompts for your FIRECRAWL_API_KEY (free tier is 500 credits/month at https://firecrawl.dev), and idempotently merges the MCP entry into ~/.claude/settings.json (existing entries preserved). See extensions/firecrawl/README.md for full setup, free-tier credit math, troubleshooting, and tool-prefix details.

Skills that gain capabilities when Firecrawl is installed: seo-page, seo-schema, seo-geo, seo-technical-audit, seo-sitemap, seo-content-audit, seo-drift, seo-content-brief, seo-competitor-pages, seo-sxo, seo-backlinks-profile. Without Firecrawl, each of these falls back to WebFetch as described above — lower-fidelity, but still a usable deliverable — and supports a --no-firecrawl flag to opt out at runtime even when the extension is installed (saves credits). Two skills are hard exceptions that need Firecrawl to run at all: the dedicated seo-firecrawl orchestrator, and seo-images — <img> attributes (srcset, loading, fetchpriority, etc.) are stripped from WebFetch's markdown, so the audit has nothing to analyse without it.

The seo-google skill needs the Google API client libraries plus a free Google Cloud project. All Google APIs used here are free (with quotas). Setup is one-time. Adapted with attribution from AgriciDaniel/claude-seo's upstream seo-google skill (MIT) — same Python scripts and reference docs, namespaced to ~/.config/seo-skills/.

# From inside a clone of this repo:
bash extensions/google/install.sh

The script verifies Python 3.10+, pip-installs the Google API libraries (google-api-python-client, google-auth*, google-analytics-data, plus matplotlib + weasyprint + openpyxl for PDF/HTML/XLSX reports), creates ~/.config/seo-skills/, writes a stub google-api.json, and runs the credential checker.

Fill in the config per skills/seo-google/references/auth-setup.md (8-step walkthrough). The skill detects which tier is unlocked and only enables the matching commands — Tier 0 (API key) gives PSI + CrUX + YouTube + NLP + Knowledge Graph + Web Risk; Tier 1 (+ service account) adds GSC + Indexing; Tier 2 (+ GA4 property ID) adds GA4; Tier 3 (+ Ads dev token) adds Keyword Planner. See extensions/google/README.md for full setup, troubleshooting, and rate-limit reference.

Skills that gain capabilities when Google APIs are configured: seo-google (the dedicated skill — fully gated on this extension), plus seo-page, seo-drift, seo-technical-audit, seo-content-audit, seo-sitemap, seo-geo, seo-keyword-cluster, seo-keyword-niche, and seo-plan (real impressions/clicks/conversions/CWV instead of estimates when configured).

Each skill is a single SKILL.md file with YAML frontmatter. Claude loads the frontmatter at startup to know when to use the skill, then loads the body only when triggered, then reads any bundled reference files only when needed. This is Anthropic's progressive disclosure pattern.

A typical skill run produces:

1. A set of numbered raw-data markdown files (01-domain-overview.md, 02-competitors.md, etc.) for auditability.
2. Optionally, a CSV for spreadsheet work.
3. A final synthesised deliverable (BRIEF.md, REPORT.md, or PLAN.md) written in a format a human reviewer or downstream stakeholder can consume directly.

The skills chain naturally. A typical agency workflow:

1. Run seo-competitor-gap-analysis to see what opportunities exist.
2. Run seo-keyword-cluster on the gap output to plan content.
3. Run seo-content-brief on each cluster's pillar topic.
4. Run seo-backlink-gap to identify link targets for the new content.
5. Run seo-ai-search-share-of-voice quarterly to track LLM visibility.

seo-skills/
├── .claude-plugin/
│   ├── marketplace.json                # Claude Code marketplace manifest
│   └── plugin.json                     # Plugin manifest
├── skills/
│   ├── seo-content-brief/
│   │   └── SKILL.md
│   ├── seo-ai-search-share-of-voice/
│   │   └── SKILL.md
│   ├── seo-backlink-gap/
│   │   └── SKILL.md
│   ├── seo-keyword-cluster/
│   │   └── SKILL.md
│   ├── seo-competitor-gap-analysis/
│   │   └── SKILL.md
│   ├── seo-agency-landing-page/
│   │   └── SKILL.md
│   ├── seo-page/
│   │   └── SKILL.md
│   ├── seo-schema/
│   │   ├── SKILL.md
│   │   ├── templates/                  # 6 JSON-LD starter templates
│   │   └── references/google-rich-results.md
│   ├── seo-drift/
│   │   ├── SKILL.md
│   │   └── references/drift-thresholds.md
│   ├── seo-sxo/
│   │   ├── SKILL.md
│   │   └── references/                 # page-type patterns + persona rubrics
│   ├── seo-technical-audit/
│   │   ├── SKILL.md
│   │   └── references/severity-mapping.md
│   ├── seo-content-audit/
│   │   ├── SKILL.md
│   │   ├── references/                 # core-eeat (60 items) + cite (30 items)
│   │   └── templates/verdict.md
│   ├── seo-sitemap/
│   │   └── SKILL.md
│   ├── seo-competitor-pages/
│   │   └── SKILL.md
│   ├── seo-backlinks-profile/
│   │   └── SKILL.md
│   ├── seo-subdomain/
│   │   └── SKILL.md
│   ├── seo-geo/
│   │   └── SKILL.md
│   ├── seo-ads/
│   │   └── SKILL.md
│   ├── seo-keyword-niche/
│   │   └── SKILL.md
│   ├── seo-firecrawl/                  # Ad-hoc Firecrawl orchestrator (extension required)
│   │   └── SKILL.md
│   ├── seo-plan/
│   │   └── SKILL.md
│   ├── seo-google/                     # Google APIs skill (extension required)
│   │   ├── SKILL.md
│   │   ├── assets/templates/           # 3 report templates (cwv-audit, gsc-performance, indexation)
│   │   └── references/                 # 10 API reference docs (auth, GSC, PSI, CrUX, GA4, etc.)
│   └── seo-api/                        # SE Ranking API integration architect
│       ├── SKILL.md
│       └── references/                 # auth-and-keys, rate-limits-and-credits, api-surface-map, integration-patterns
├── scripts/                            # Python scripts called by seo-google (forked from AgriciDaniel/claude-seo, MIT)
│   ├── google_auth.py
│   ├── pagespeed_check.py
│   ├── crux_history.py
│   ├── gsc_query.py
│   ├── gsc_inspect.py
│   ├── indexing_notify.py
│   ├── ga4_report.py
│   ├── youtube_search.py
│   ├── nlp_analyze.py
│   ├── keyword_planner.py
│   └── google_report.py
├── extensions/
│   ├── firecrawl/                      # Optional Firecrawl MCP wiring
│   │   ├── install.sh
│   │   ├── uninstall.sh
│   │   └── README.md
│   └── google/                         # Optional Google APIs Python toolchain
│       ├── install.sh
│       ├── uninstall.sh
│       ├── README.md
│       └── LICENSE-AgriciDaniel.txt    # MIT attribution for forked content
├── examples/                           # Real, end-to-end runs against public targets
│   └── seo-ai-search-share-of-voice-wix-com-20260427/
├── install.sh                          # Top-level installer (clone + extension installers)
├── CHANGELOG.md
├── LICENSE
└── README.md

The repo is a Claude Code marketplace whose single plugin is the repo itself (marketplace plugins[0].source is "./"). This keeps skill folders visible at the top level rather than buried two layers deep.

Every skill writes its output to a folder named {skill-slug}-{target-slug}-{YYYYMMDD}/ (e.g., seo-content-brief-example-com-20260427/). The date stamp keeps re-runs non-destructive and makes it easy to diff outputs over time. Browse examples/ to see what a finished run looks like before installing.

The SE Ranking MCP server has two namespaces with different rate limits:

• Data API (DATA_*): 10 requests per second. Charges API credits per call.
• Project API (PROJECT_*): 5 requests per second. Retrieval and management calls do not charge; writes consume quota slots.

Every skill in this repo is designed to pace sequentially inside these limits. The largest skills (seo-competitor-gap-analysis with full keyword dumps, seo-keyword-cluster with 20 seeds) can consume thousands of credits on large domains. Check DATA_getCreditBalance before running on production accounts, and use the ceiling parameters the skills document to cap cost.

This repo is the official Skills catalogue for the SE Ranking MCP server. See CONTRIBUTING.md for the skill-authoring template, required structure, MCP tool conventions, and PR checklist.

• Bug reports — open an issue with the skill name, the input, and the unexpected output.
• New skills — PRs welcome. Use the template in CONTRIBUTING.md.
• Improvements — PRs welcome. Include a before/after example run in the PR description.

• SE Ranking: https://seranking.com
• SE Ranking API: https://seranking.com/api.html
• SE Ranking remote MCP: https://seranking.com/api/integrations/mcp
• Claude Agent Skills: https://platform.claude.com/docs/en/agents-and-tools/agent-skills/overview
• Claude Code Plugins: https://code.claude.com/docs/en/plugins

MIT. See LICENSE.