What is seo-skills?
seo-skills is a collection of 30 Agent Skills for search engine optimization, visibility in AI answer engines (GEO), link building, content and copywriting. Each skill is a folder with one SKILL.md file that follows the Agent Skills specification: a short name and description in the frontmatter, then decision trees, thresholds, output file paths and a done checklist. The coding agent reads only the name and description at startup and loads the full file when a prompt matches. The skills are written for Claude Code, Cursor, OpenAI Codex, OpenCode, Windsurf, Cline and Roo Code, and the README describes them as workflows that "end in a merged PR or a file in your repo, not an audit PDF". Most skills edit files in the user's repository or write Markdown, CSV and JSON files under seo/ and marketing/.
The project is written and maintained by Ilias Ism, who publishes it as Illyism on GitHub and runs the companion catalog at il.ly/skills. His about page describes him as a Belgian software engineer, SaaS founder and search strategist, a co-founder of the HR platform Officient (acquired by Exact in 2020), and the creator of LinkDR, GenPPT, SEO Roast and the seo-skills library. The repository was created on June 10, 2026 with two skills, blitz-seo and blitz-copy, and grew over the following months. In October 2026 the engineering skills (eval and extension-store-assets) moved to the separate Illyism/dev-skills repository, so the il.ly/skills catalog lists 36 skills: the 30 in this repository plus code and product skills such as eval that are installed from dev-skills.
There are no GitHub releases or tags. The version is carried in the plugin manifests .claude-plugin/plugin.json, .cursor-plugin/plugin.json and gemini-extension.json, which all state 1.3.0; each SKILL.md also has its own metadata version, currently 1.0.0. Version 1.3.0 was committed on November 12, 2026 and expanded the set to 30 skills with the Ahrefs, DataForSEO, proof-kb, seo-roast, bottom-of-funnel and distribution workflows. Later commits added a check for poisoned Google results to dataforseo-serp (November 16), replaced the README visuals and tightened the skill introductions (November 20). Installation goes through the npx skills add Illyism/seo-skills installer, and the repository also ships an .mcp.json file that preconfigures three hosted MCP servers.
The README's License section reads "MIT © Ilias Ism", the plugin manifests declare "license": "MIT", and the il.ly/skills page links the repository as MIT. The repository does not contain a separate LICENSE file with the license text, so GitHub's license detection shows no license for it. The repository description calls the collection "30 open-source SEO skills". Besides Markdown, the repository holds two helper scripts: a Bash redirect checker and a Node.js link graph script.
Several skills are built around services that the author runs or that sponsor the repository: aiseotracker, ai-seo-report and content-coverage use AI SEO Tracker, linkdr plans orders for the managed backlink service LinkDR, awesome-tools submits products to the Awesome Tools directory and scribiz uses the Scribiz transcription service. The disclosure in the launch article states that these paid products belong to the author and sponsor the repository, and that 24 of the 30 skills run without any of them. The skills carry no data of their own: live SERP, keyword and backlink data comes from the user's own Ahrefs or DataForSEO access, Search Console or web search, and the project operates no hosted service for the skills themselves.
Key capabilities
- 30 skills in four folders The repository groups the skills as 20 SEO skills in
seo/, 5 content skills incontent/, 3 marketing skills inmarketing/and 2 copywriting skills incopywriting/. The README sorts them by job instead: 12 technical, research and on-page skills, 5 AI search (GEO) skills, 4 link and directory skills, 7 content and distribution skills and 2 copywriting skills. The README gives each skill's output and an example prompt, and il.ly/skills shows every skill with a copy button. - Installer for all or single skills The documented install is
npx skills add Illyism/seo-skillsfor all 30 skills,--skill seo-autopilot(or any other name) for one skill and--listto show what is available. Claude Code reads skills from.claude/skills/in a project or from~/.claude/skills/, and the README's command for it isnpx skills add Illyism/seo-skills -a claude-code; Cursor uses.cursor/skills/or.agents/skills/and the-a cursorflag. Codex, OpenCode, Windsurf, Cline and Roo Code read the same files from.agents/skills/. - Copy and paste for chat assistants For ChatGPT, Claude.ai, Perplexity and Grok the README says to copy a skill from il.ly/skills into the chat as Markdown. The launch article notes that a chat window has no repository access, so the result there is a plan or a draft that the user applies by hand rather than an edited file or pull request.
- Plugin manifests and MCP configuration The repository contains
.claude-plugin/plugin.jsonand.cursor-plugin/plugin.json, which point at the four skill folders and at.mcp.json, plus agemini-extension.jsonwith the same MCP servers..mcp.jsonpreconfigures three hosted HTTP MCP servers: AI SEO Tracker (https://aiseotracker.com/api/agent/mcp), Scribiz (https://scribiz.com/mcp) and Awesome Tools (https://awesome.tools/mcp). - seo-autopilot for daily runs
seo-autopilotis a prompt meant to run on a daily schedule in a marketing site's repository after five lines are filled in: site, product, conversion, competitors and ship mode (pull request for review, or merge when the build passes). Because the agent has no memory between runs, all state lives inseo/:context.md,keyword-map.csv,backlog.md,log.md,ai-prompts.mdand competitor sitemap snapshots. It uses whatever data sources are connected (Search Console, Ahrefs or a similar tool, analytics, web search) and is instructed to report a missing source instead of guessing. - seo-roast teardown that ends in a pull request
seo-roastsorts findings into four tiers: fatal (blocks crawling, indexing or rendering of a revenue page), revenue, compounding (authority, links, AI visibility) and fluff such as alt text or meta length. Fatal issues are fixed in the same pull request before anything else, and the rest goes intoseo/roast-roadmap.mdordered by expected revenue, with an optional client deck. - Ahrefs research with three connection paths
ahrefs-researchfirst checks for Ahrefs MCP tools, then for anAHREFS_API_KEYto call API v3, and otherwise asks the user for a CSV export from the Ahrefs interface. Its workflows write a competitor top-pages gap list (seo/competitor-gaps.md) and scaffold the top gap pages, filter link prospects with inflated Domain Rating intoseo/link-targets.csv, mine subreddit topics and pick pages for link insertions. An Ahrefs seat without an API plan is enough for the CSV path. - DataForSEO SERP workflows
dataforseo-serpreads live Google results, keyword volume and CPC, AI Overview references and LLM answer citations through the DataForSEO MCP tools or REST v3 withDATAFORSEO_LOGINandDATAFORSEO_PASSWORD. It clusters keywords by overlap of the top-10 URLs intoseo/keyword-clusters.json, filters free tool ideas by CPC and volume, and writes a per-page intent lock before any draft. Without credentials it falls back to web search for fewer than 10 queries and stops to ask for credentials at 10 or more. - proof-kb first-hand fact store
proof-kbbuilds a knowledge base inseo/kb/with brand facts and proof cards for benchmarks and customer results, which drafts and page upgrades pull from. Itsroute-factstep finds pages that quote an old number after a benchmark or price change and rewrites them. - Technical changes with helper scripts
domain-migrationhandles rebrands and subdomain-to-subfolder moves with a URL and backlink inventory, a one-hop 301 redirect map,scripts/verify-redirects.shand backlink update email drafts.internal-linkingshipsscripts/internal-link-graph.mjsfor an inbound link audit and adds body links from high-traffic posts to pricing, comparison and tool pages.block-ai-scrapersbuilds a gate for Cloudflare Workers, Next.js middleware, Hono, Express or Fastify that blocks AI agents, CLI scrapers and headless browsers while letting Googlebot through only after IP and ASN verification, not by user agent. - On-page and page-level skills
seo-title-optimizationrewrites title tags, frontmatter titles and H1s against the live results and truncation limits,blog-refreshercompares a slipping post with the current top 10 and rewrites weak sections,blitz-seoplans a 30-day sprint on one page and one money keyword, andvs-and-alternativesbuilds/vs/[competitor], alternative and persona pages with sourced switching reasons, seat-based pricing tables and migration steps. - AI search (GEO) skills
customer-journeymaps a five-stage awareness journey with the Google keywords and AI assistant prompts buyers use at each stage.aiseotracker,ai-seo-reportandcontent-coverageconnect to AI SEO Tracker over MCP with browser OAuth instead of an API key and read prompt rankings, cited URLs and competitors across ChatGPT, Perplexity, Gemini, Copilot and Google AI Overviews; according to the skill, AI SEO Tracker sells scans as a one-time audit or as weekly plans.reddit-serpwritesseo/reddit-serp-queue.mdwith unlocked Reddit threads that rank on page 1 or are cited by AI answers, plus reply drafts and subreddit rule notes. - Link building and directories
directory-submissionsbuilds onesubmission-manifest.jsonfrom the codebase, works through 50 directories in five waves, logs attempts insubmissions-log.jsonand lists the repository files that MCP and agent registries index.honest-link-outreachwrites a dated outreach batch with a SEND/CUT prospect table and plain-text emails.linkdrplans a LinkDR order by Domain Rating tier and returns one checkout link for the user to approve, andawesome-toolssubmits a product to Awesome Tools over its JSON API or browser WebMCP tools, with a free listing path that has link and traffic conditions and paid listing options. - Content editing and images
unslop-seoedits Markdown, MDX, TSX or HTML files in place in three phases (ground claims in facts, restructure for answer engines with the answer in the first 40 words, remove AI writing patterns) and includes a check script that must report 0 hits.seo-blog-imagesplans and compresses cover images, Playwright screenshots and SVG charts withsharp; generated images go through OpenRouter (OPENROUTER_API_KEY, default modelgoogle/gemini-2.5-flash-image) or the agent's built-in image tool, and the skill forbids AI-drawn product UIs and charts. - Video, distribution and changelog content
youtube-seoturns a recording or.srtfile into two title options, a chaptered description, launch posts and aVideoObjectembed.scribizgets transcripts, on-screen text and chapters through Scribiz, with a small daily allowance without a key and the account's minutes withSCRIBIZ_API_KEY.parasite-and-prwrites a press release, a canonicalized Dev.to draft and a LinkedIn article intomarketing/distribution/<slug>/without submitting anything,diff-marketingdrafts changelog entries and launch posts from recent commits, andintrowrites blog intros by search intent. - Copywriting skills
understand-customersproduces a buyer research brief with pains, failed alternatives, objections and the customer's own phrasing before any copy is written.blitz-copywrites landing page sections, cold emails, social hooks and SERP titles with the Hook, Spike and Payoff formula described in the skill. - Documented skill chains The README lists six end-to-end workflows that chain skills: a post with first-hand proof, pages for buyers leaving a competitor, pushing a page-2 URL into the top 3, getting cited by AI assistants, launching a new domain from DR 0, and daily repository SEO with
seo-autopilotplus a weeklydiff-marketingrun. - Requirements and third-party costs The skills need an agent that reads
SKILL.mdfiles and, for most workflows, a local checkout of the site's repository; the installer runs throughnpx. Optional keys are bring-your-own:AHREFS_API_KEY,DATAFORSEO_LOGINandDATAFORSEO_PASSWORD,OPENROUTER_API_KEYandSCRIBIZ_API_KEY, and Search Console data is used when the agent has it connected. Accounts, API usage and any LinkDR, AI SEO Tracker or Awesome Tools purchases are paid by the user to those vendors; model tokens go through the user's own agent subscription.
Where seo-skills fits
Fixing crawl and indexing blockers in a site repository
The README's example prompt is "Run seo-roast on our repo and fix the fatal blockers in a PR". The skill opens with a one-line verdict, checks crawler access in robots.txt, rendering of revenue pages, programmatic page templates and whether pricing appears as plain text in the HTML, then fixes the fatal tier in one pull request and writes the remaining items to seo/roast-roadmap.md. The launch article shows examples such as a Next.js robots.ts change that unblocks Bing and AI crawlers and a server-rendered pricing table.
Running SEO as a scheduled agent job
seo-autopilot is designed to run daily in the marketing site's repository. Each run reads the state files in seo/, requests every URL in the sitemap and fixes what broke, compares recent commits with pages that still state an old price or limit, and ships one backlog item as a pull request or a merge, depending on the ship mode. Weekly it works queries at positions 8 to 20 and writes a short report; outreach emails and directory copy are drafted but never sent.
Publishing a post backed by first-hand proof
The first documented chain starts with dataforseo-serp to cluster keywords and lock the search intent, adds benchmarks and customer results with proof-kb, creates images with seo-blog-images, cleans the draft with unslop-seo, adds internal links with internal-linking and prepares syndication copies with parasite-and-pr. Each step writes files into the repository, such as seo/keyword-clusters.json, proof cards in seo/kb/ and the distribution bundle.
Comparison and alternative pages for competitor searches
For buyers leaving a competitor, the README chains ahrefs-research to diff the competitor's top pages, vs-and-alternatives to build /vs/ and alternative routes with pricing tables for 1, 5 and 25 seats, reddit-serp to queue ranking Reddit threads and honest-link-outreach for link requests. The Ahrefs step works through the MCP server, API v3 or a CSV export from the Ahrefs interface.
Tracking and closing gaps in AI assistant answers
The AI search chain runs customer-journey to list the prompts buyers ask at each awareness stage, connects AI SEO Tracker through aiseotracker, summarizes cited sources and changes with ai-seo-report and plans posts for missing prompts with content-coverage. These three skills depend on an AI SEO Tracker account, which is a paid service run by the author; customer-journey works without it.
Launching a new domain and moving an existing one
For a new product the README chains understand-customers, blitz-copy, directory-submissions and awesome-tools, then blitz-seo once the domain has some authority. For an existing site that changes domain or moves blog.example.com into a subfolder, domain-migration builds the URL inventory from the old sitemap, the route tree, Search Console, a backlink tool and analytics, writes a one-hop redirect map and runs scripts/verify-redirects.sh, which must report 0 failures. Similar skill collections listed on iThub include Claude SEO, GEO-SEO Claude, NotFair Plugin and avoid-ai-writing.
Install
npx skills add Illyism/seo-skillsReviews of seo-skills
Be the first to share your experience with seo-skills.
Write a review
Loading review form…
Found seo-skills useful? Upvote it todayYour vote helps other IT teams find it in the daily ranking.



