SuperSEO Skills
11 opinionated Claude skills for SEO — page audits, briefs, articles, E-E-A-T scoring, and link plans — one input each, and the agent does the research itself.
Evidence: SECURITY.md declares no executable code; skills only fetch user-supplied URLs/keywords; the only persistence (write-content saving business context) requires explicit user-designated location — good least-privilege and data-flow transparency posture. Deductions: these are claims, not verified file contents — no SKILL.md was provided, so internal instructions cannot be confirmed to match; user confirmation is mentioned only once; no rate/consent constraints on external fetches against third-party sites; no rollback/uninstall guidance; business context saved by write-content is sensitive but storage/protection details are absent. License and attribution are complete.
Evidence: README describes a consistent single pattern across 11 skills; a demo cast and multi-environment install paths are given. Deductions: skills themselves absent from evidence, so internal consistency is unverifiable; reliance on agent runtime web-fetch/search means SERP availability is realistically fragile and undiscussed; 404/failure handling is claimed only in a comparison table with no documented failure messages or degradation behavior.
Evidence: audience (SEO practitioners/agencies) and scenario coverage are explicit; capability boundaries are honest — skills cannot access GSC history, commercial platform covers the rest; five environments (Claude Code/Claude.ai/Desktop/Cursor/any agent) each get concrete adaptation instructions. Deductions: trigger precision rests on Claude skill discovery with no trigger-word/description metadata available for review; Cursor degradation is acknowledged but only in prose.
Evidence: install notes are exceptionally thorough (three-command plugin install, manual clone, symlink, zip upload, Cursor loop); Apache-2.0 full text included with an explanation; standard Claude skill directory layout; limitations (no GSC access, no cross-surface sync) explicitly listed; CONTRIBUTING and SECURITY processes exist with response SLAs (72h/2 weeks). Deductions: no CHANGELOG file or versioning scheme in evidence (only an example tag v0.2.0); examples are a single demo GIF/cast with no FAQ; naming stability unverifiable across the missing skill files.
Evidence: clear differentiation from paste-data prompts; marginal value is agent-side competitor research and structured output at near-zero runtime cost (Markdown only). Deductions: output usability cannot be verified — no SKILL.md or sample output is in evidence; claims like 1-10 scores and char-counted rewrites are unproven; the demo cast cannot be checked in a static review.
Evidence: README separates factual claims from marketing comparison (the table is framed as contrast narrative); named methodology sources (Koray, Roof, Ray) are traceable in principle; the SECURITY.md scope matches the workflow code (zip packaging only). Deductions: core quality claims ('production-tested', 'realistic conversion rates from the field', CTR benchmarks) have no traceable data or citations; no third-party corroboration; references/ contents were not provided, so methodology claims cannot be checked.
- This review is based only on README/LICENSE/SECURITY.md and the release workflow; no SKILL.md or references/ files were available, so all claims about skill behavior, quality, and safety boundaries are file-level unverified.
- Skills direct the agent to search Google and fetch target sites; the repo imposes no constraints on rate, robots/ToS compliance, or consent of the fetched sites.
- write-content persists user business context into CLAUDE.md or similar; review what gets written for sensitive business information.
- The README contains overt marketing content (the comparison table, 'production-tested'); treat these as author claims, not verified facts.
- Publisher identity is unverified in the FollowAgents registry; assess trust independently before adoption.
What does this agent do, and when should you use it?
SuperSEO Skills is an open-source collection of 11 Claude skills from InhouseSEO, distributed as Markdown skill files with on-demand reference libraries, covering page audits, content briefs, article writing, keyword deep-dives, semantic gap analysis, E-E-A-T audits, topic cluster planning, featured snippet optimization, link building, and expert interviews. Unlike prompts that require pasting keyword exports, each skill accepts a single URL or keyword and has the agent fetch the page, run Google searches, and read the top-ranking competitors itself. The methodology draws on Koray Tuğberk, Kyle Roof, and Lily Ray, plus a generation-time anti-AI-slop ruleset with tiered banned vocabulary, structural pattern detection, and the Horoscope Test. The skills are production-tested at InhouseSEO, licensed under Apache 2.0, and contain no executable code — the security surface is limited to Markdown. Historical Search Console analysis and site monitoring are out of scope and require the companion commercial platform InhouseSEO.
After installation, each skill takes one input and drives the agent through a full workflow: page-audit fetches the target URL, identifies the primary keyword, runs a Google search, reads the top 3 competitors, and returns a 7-dimension scored audit; content-brief Googles the keyword, reads the top 10 results, maps intent, and produces a writer-ready brief with SERP gap analysis; keyword-deep-dive classifies SERP intent and outputs a 90-day ranking plan; write-content asks 2-3 questions about your expertise, then writes the article under the full anti-AI-slop ruleset; eeat-audit scores Experience, Expertise, Authoritativeness, and Trustworthiness with specific fixes including YMYL scoring; semantic-gap-analysis lists the exact entities and subtopics missing versus the top 3 (with EAV triple worked examples); linkbuilding applies phase-appropriate tactics (foundation, growth, authority) across 9 playbooks, with anchor text safety and link velocity red-flag references. All skills live at skills/<name>/SKILL.md with bundled references/ folders containing 23 content-type templates, writing technique modules, and diagnostic benchmarks.
- A site owner with an underperforming landing page URL who wants a competitor-benchmarked, char-count-rewritten 7-dimension audit within minutes.
- A content team handed a target keyword that needs an intent-mapped, SERP-gap-analyzed brief ready to hand to a writer.
- An SEO specialist who wants to know exactly which entities and subtopics their page is missing versus the top 3 results, using semantic-gap-analysis.
- An agency that needs to batch-score client pages for E-E-A-T and receive concrete, actionable fixes (including YMYL scoring).
- A new-site owner in the link-building foundation phase who needs phase-appropriate tactics (citations, resource pages, guest posting) plus anchor text and link velocity safety guidance.
- An enterprise content team that wants to extract first-party expert knowledge via the expert-interview question bank before running the writing workflow.
What are this agent's strengths and limitations?
- Unlike prompts that ask you to paste keyword exports, crawl reports, and backlink profiles, the agent performs competitor fetching and SERP reading itself — and catches 404/dead-link cases — with a single input.
- The methodology is explicit (POP test, Information Gain, E-E-A-T, semantic networks), and write-content embeds concrete anti-slop mechanics: tiered banned vocabulary, structural pattern detection, and the Horoscope Test.
- All 11 skills share a uniform skills/<name>/SKILL.md plus references/ layout that works identically in Claude Code, Claude Desktop, Claude.ai, and Cursor; the repo contains no executable code and ships under Apache 2.0 for commercial use.
- Results depend on the underlying model's ability to actually fetch web pages and run searches; the README offers no independent cross-model evaluation, and in Cursor references cannot be loaded dynamically and must be pasted manually.
- The skills cannot access historical Search Console data or monitor a site over time — the GSC warehouse, nightly signals, crawler, and 13 GSC-backed workflows require purchasing the InhouseSEO platform (14-day free trial), a clear commercial funnel.
- Claude.ai skill uploads require a Pro plan or above, are one-at-a-time, and don't sync across devices, creating real friction for multi-device use.
How do you install or deploy this agent?
Recommended (Claude Code plugin marketplace): run /plugin marketplace add inhouseseo/superseo-skills, then /plugin install superseo@superseo-skills; update with /plugin marketplace update superseo-skills. Manual: git clone https://github.com/inhouseseo/superseo-skills.git, then cp -r superseo-skills/skills/* ~/.claude/skills/ and restart Claude Code; alternatively symlink with ln -s ~/superseo-skills/skills/*/ ~/.claude/skills/ to keep git pull live. Claude.ai / Claude Desktop / mobile: upload skill zips one at a time via Settings → Features → Skills → + (Pro plan or above; skills don't sync across devices). Cursor: copy each SKILL.md into your .cursor/rules/ directory. No API keys or data exports required.
How do you use this agent?
In Claude Code, invoke a skill like "run page-audit on <your URL>". Each skill takes one input: page-audit takes a URL; content-brief, keyword-deep-dive, semantic-gap-analysis, and featured-snippet-optimizer take a keyword; write-content takes a topic (and will ask 2-3 questions about your expertise first); improve-content takes an existing page. The agent fetches pages, runs searches, and reads competitors on its own; when a step requires deeper material, it loads the bundled references/ files on demand. In static-rule environments like Cursor, paste the required references/ file into the conversation manually.