SuperSEO Skills: 11 Claude Skills for SEO Work, and Where They Stop

What SuperSEO Skills solves, and who it is built for

Most SEO prompt packs assume you have already done the data collection. You export keyword data, pull a crawl report, download a backlink profile, then paste it into a prompt that analyses whatever you fed it. SuperSEO Skills inverts that. The README describes the pattern as one input per skill: a URL or a keyword. The agent fetches the page, identifies the primary keyword, runs a Google search, reads the top three competitors and returns the audit. Nothing is exported and no API keys are configured.

The intended user is an SEO or content person who already runs Claude Code or a comparable agent surface and wants the research step handled inside the same session. The repository names 11 skills: page-audit, content-brief, write-content, improve-content, keyword-deep-dive, semantic-gap-analysis, eeat-audit, topic-cluster-planning, featured-snippet-optimizer, linkbuilding and expert-interview. The README states the methodology draws on Koray Tuğberk, Kyle Roof and Lily Ray, and that the skills were production-tested at InhouseSEO.

That framing matters for expectations. This is not a crawler, not a rank tracker and not a dashboard. It is a set of opinionated instructions plus bundled reference material that an agent reads on demand. If your workflow depends on scheduled crawls or historical GSC data, the skills do not replace that layer; they sit on top of a live fetch.

How a skill actually works: fetch, read, score

The mechanism is the same across the library, and the README uses page-audit as the worked example. You supply a URL. The agent fetches the page, identifies the primary keyword, runs a Google search, reads the top three competitors end to end, and produces a seven-dimension audit. The stated output includes named competitors, character-counted rewrites and a 1 to 10 score per dimension.

The README draws a direct contrast with a vanilla prompt. Where a plain prompt might hallucinate word counts, the skill fetches and reads the top three. Where a prompt might audit a 404 error page, the skill catches the error and finds the canonical. That second behaviour is the more interesting one, because it implies the agent checks the fetch result before analysing it rather than treating any response as the page.

Other skills follow the same shape with different depth. content-brief takes a keyword, reads the top 10 results, maps intent, identifies the content gap and returns a writer-ready brief. keyword-deep-dive reads the SERP, classifies intent, reads the top three competitors in detail and produces a 90-day ranking plan. write-content researches, asks two or three questions about your expertise, then writes using the anti-AI-slop ruleset.

The reference layer is where the library carries weight. Each skill ships its own references folder. The README lists 23 content-type templates distributed across write-content, improve-content, content-brief, eeat-audit and featured-snippet-optimizer, each with H1 and H2 structure, schema markup, snippet strategy, word count targets and E-E-A-T signals. Diagnostic references are skill-specific: the POP test hierarchy in page-audit, EAV triple worked examples in semantic-gap-analysis, YMYL scoring in eeat-audit, snippet format templates in featured-snippet-optimizer.

Installing SuperSEO Skills and running a first page audit

The README gives a two-command install for Claude Code through the plugin marketplace. Run both from your Claude Code session, not from a shell in the repository.

/plugin marketplace add inhouseseo/superseo-skills
/plugin install superseo@superseo-skills

After the second command completes, the skills are available in the session. The README's instruction for a first run is a single line in Claude, with your own URL substituted for the placeholder.

run page-audit on <your URL>

The demo material in the repository shows this running on backlinko.com/title-tags and reports a runtime of 1m 3s. Expect the agent to fetch the page, search for the primary keyword and read the top three competitors before it writes anything. What you should see in the output is a seven-dimension audit with named competitors, rewrites with character counts, and a score from 1 to 10 per dimension.

If you are not in Claude Code, the README points to a Quick start section for Claude Desktop, Claude.ai, Cursor and other agent surfaces. That section is truncated in the README available here, so treat the plugin route as the documented path and check the repository README for the manual route before assuming a specific directory layout. The skills live under skills/ in the repository, with a .claude-plugin/ directory at the top level for the plugin manifest.

The anti-AI-slop ruleset and what it constrains

The generation-time ruleset is the part of this repository with the clearest editorial opinion. The README describes tiered banned vocabulary, structural pattern detection and something called the Horoscope Test, applied inside write-content, improve-content and expert-interview. The intent is to stop the agent producing the flat, hedged, list-shaped prose that SEO teams keep having to rewrite.

Structural pattern detection is the more consequential half. A banned word list is easy to satisfy by swapping synonyms. Detecting repeated sentence shapes and list-shaped paragraphs is harder, and it is the failure mode that makes generated content recognisable regardless of vocabulary. The README does not publish the rules themselves, so you cannot audit the list before installing; you see its effect in the output or not at all.

There is a tension worth naming. write-content asks you two or three questions about your expertise before writing. That is the mechanism for injecting first-party experience, and it is also a hard dependency: if you have nothing to say about the topic, the skill has no material to work with and the ruleset will not manufacture it. expert-interview exists for the same reason, extracting expertise through targeted questions and feeding it into content writing. Treat the pair as a research step, not a formatting step.

The ruleset also constrains the agent's own output. An audit that cannot use certain constructions is more likely to state a number than to hedge around one, which is the point. Whether the tiered list is calibrated to your house style is something only your own review will tell you.

Where SuperSEO Skills is the wrong tool

The library is a set of instructions for an agent that can fetch pages and run searches. That is a real constraint, not a footnote. If your environment blocks outbound fetches, or the agent cannot run a search, most of the skills have nothing to work from. The README's whole pitch is that the agent does the data gathering itself, so removing that capability removes the product.

The second boundary is scale. Every skill takes one input: one URL, one keyword, one topic. There is no documented batch mode, no queue, no scheduling and no crawl. If you need to audit 400 pages, you are running the skill 400 times and reading 400 outputs. The README does not describe a way around that.

The third is the link-building coverage. The repository ships nine tactic playbooks under skills/linkbuilding/references/tactics/, covering entity stacking, citations and directories, competitor backlink gap, guest posting, resource pages, skyscraper, strategic partnerships, podcast guesting and a new-site launch strategy. The README states plainly that nine further playbooks, including link reclamation, broken link building, digital PR, linkable assets, niche edits and ego bait, live in the InhouseSEO platform rather than this repository. If those are the tactics you need, the open repository is not the complete set.

Finally, the README does not document rollback, offline use or a self-hosted mirror. It is silent on all three. If your review process requires being able to pin and revert a prompt library, you will need to establish that yourself from the repository history.

How this differs from generic SEO prompt packs

The obvious alternative is a general-purpose SEO prompt collection, the kind that asks you to paste in keyword exports and crawl data and then reformats them. The difference is where the data comes from. A paste-in prompt is deterministic about its inputs and useless without them; a SuperSEO skill is non-deterministic about its inputs and useless without network access. Neither is strictly better. If you already have clean exports and want a repeatable transformation, the paste-in approach gives you the same answer twice. If you want the research done, you are trading reproducibility for coverage.

The second alternative is a conventional SEO suite with a crawler, a rank tracker and a backlink index. Those tools answer questions about your whole site over time. SuperSEO Skills answers questions about one page or one keyword right now. The overlap is the audit and the brief, and even there the outputs differ: a suite gives you a table of issues, the skill gives you named competitors and character-counted rewrites.

A third comparison is writing directly in Claude without any skill. The README's own table frames this as the baseline, and the honest read is that a careful operator with a good prompt can get close on a single page. The library's advantage is consistency across 11 tasks and the bundled references, which is a maintenance argument rather than a capability argument. If you only ever write articles, one well-tuned prompt may be enough. If you move between audits, briefs, E-E-A-T reviews and link plans, having the reference material bundled per skill is the thing you are actually installing.

Licence, maintenance and upgrade cost

The repository is licensed Apache-2.0, with a LICENSE file and a NOTICE file at the top level. Apache-2.0 permits commercial use and modification and includes an explicit patent grant. It also requires that you preserve the licence and NOTICE files and state significant changes when you redistribute. For a prompt and reference library that you might fork and adapt internally, that is a permissive setup, but the NOTICE obligation is easy to forget when the artefacts are markdown rather than compiled code. Nothing here is legal advice; read the LICENSE and NOTICE yourself before redistributing a modified copy.

On maintenance, the last push to the default branch was on 2026-09-03, and the most recent release is v0.2.1 on the same date, described as grounding rules and autonomous-run support. The prior release, v0.2.0, was on 2026-04-14 and added per-skill references and the plugin marketplace. The gap between those two dates is roughly five months, and the repository is not archived. Two releases in that window is a slow cadence rather than a dormant one, and the content of v0.2.0 explains why: per-skill references mean the substantial work sits in files that change without a version bump.

Upgrade cost is low by construction. The install is two plugin commands and the artefacts are markdown, so pulling a new version does not require rebuilding anything. The real cost is re-reading outputs after a change to the anti-AI-slop ruleset or the reference templates, because those are the parts that alter generated prose. If you fork the skills to match your house style, you take on merging upstream changes yourself, and the repository does not document a merge path.

Editorial conclusion

Adopt SuperSEO Skills if you already work inside Claude Code or another agent surface and want page audits, briefs, E-E-A-T reviews and link plans that pull their own SERP data from a single URL or keyword. Skip it if you need bulk crawling, GSC export analysis or a hosted dashboard, since the repo is a skill library and the README points to the InhouseSEO platform for the remaining nine link-building playbooks. Before relying on it, run page-audit on one of your own URLs and check the named competitors, the char-counted rewrites and the 1 to 10 scores against what you know about that page.

Frequently asked questions

Official sources

1. inhouseseo/superseo-skills on GitHub
2. License: Apache-2.0
3. Project website
4. README
5. Releases