Introduction
If your DevTool, AI-agent platform, or B2B SaaS product is not showing up in ChatGPT, Perplexity, or Google AI Overviews, your documentation has an AI discoverability problem, not just a ranking problem.
The rules of search changed. Traditional SEO gets your content indexed by Google crawlers. Generative Engine Optimization (GEO) gets your content cited, parsed, and recommended by the large language models that now answer a growing share of technical search queries. The two disciplines overlap but require different checks, different signals, and different tooling.
Claude skills solve exactly this problem. These are free, open-source reusable instruction packages that teach Claude how to run a specific workflow inside Claude Code or Claude.ai, from technical documentation audits to full SEO growth reports, with zero dashboards, zero logins, and zero configuration overhead for most of them.
This guide covers what Claude skills are, how the leading collections are structured, which skill packs are most relevant for SEO, GEO, and developer marketing teams, and how to get started in under ten minutes.
What Are Claude Skills?
A Claude skill is a self-contained instruction package. At its core, each skill is a SKILL.md file that tells Claude when to activate, what the skill is designed to do, and how to execute the workflow step by step. Most skill packs also include supporting scripts, reference data, and a full documentation layer.
When you clone a skills repository and open Claude Code, skills load automatically. You activate them using slash commands, such as /dev-gtm docs-audit for a documentation audit or /dev-gtm growth-report for an SEO performance report.
The result is a repeatable, opinionated workflow that runs inside your terminal without leaving a browser tab open, without sending your data to a third-party dashboard, and without requiring you to map the methodology yourself.
Skills are different from prompts. A prompt gives Claude a one-time instruction. A skill gives Claude a persistent, specialized capability that it can invoke every time the trigger condition is met, following the same methodology, producing the same structured output, and integrating with external data sources like Ahrefs or Google Search Console when relevant.
The open-source Claude skills for developer marketing teams available on GitHub ship five curated skill packs, each targeting a specific team.
Why SEO and GEO Require Different Tooling in 2026
Traditional SEO tools, from Ahrefs to Screaming Frog to Semrush, were built to optimize content for Google's crawling and indexing pipeline. They check meta tags, heading structure, internal links, page speed, and backlink profiles. These signals still matter.
But AI search engines do not rank pages. They cite sources. When a developer asks ChatGPT how to instrument their application with OpenTelemetry, the model does not return ten blue links. It synthesizes an answer and attributes it to the sources it found most citable.
Citability is not the same as rankability. A page can rank in position one on Google and still never appear in an AI-generated answer because the documentation does not meet the structural and semantic signals that LLMs use to evaluate trustworthiness and relevance.
The signals GEO cares about include:
- Presence and structure of an llms.txt file
- Logical heading hierarchy and section completeness
- Breadcrumb navigation and canonical URL clarity
- Internal linking depth and documentation interconnectedness
- Install commands with version pinning
- Troubleshooting indexes and error documentation
- Schema markup for structured entity recognition
- Content freshness signals (pages under 30 days old receive significantly more AI citations)
Traditional SEO audits do not check any of these. This is why developer teams with solid Google rankings still find themselves invisible in AI search.
Claude skills bridge this gap by running both layers simultaneously, auditing your documentation against traditional on-page SEO signals and AI citability signals in the same pass, scored and reported together.
The Five Claude Skill Packs for Technical Teams
The Claude Code skills for developer GTM repository organizes skills into five packs, each built for a distinct team function.
1. Marketing Team Skills
This pack covers the workflows marketing teams run most often: SEO audits, GEO and AI visibility scoring, content brief generation, and distribution workflows. Each skill runs from a single command inside Claude Code with no external dependencies for the core checks.
The SEO audit skill runs a full technical and on-page check, identifies orphan pages with no internal links, surfaces pages with zero outgoing links (dead-end pages), and generates a prioritized fix list ordered by traffic impact.
The GEO module evaluates AI citability across the signals listed above, including llms.txt presence, structured content, internal linking depth, and documentation completeness.
2. Writing Team Skills
The writing pack covers long-form drafting, editing, style and tone enforcement, and content repurposing. Skills in this pack include a blog audit tool that flags vague language, abstract nouns, filler words, and marketing tone, with suggested rewrites for every flagged sentence, a content brief generator, and a humanizer that rewrites AI-generated text to match platform-native voice.
For teams producing technical blog content, this pack handles the full lifecycle from keyword brief to publication-ready draft.
3. Developer GTM Skills
This pack is purpose-built for developer go-to-market execution. Skills cover technical documentation auditing, API documentation quality scoring, SDK documentation review, quickstart guide validation, and developer-first content workflows.
The docs audit skill runs 33 checks across your documentation, scoring it from 0 to 100 on AI discoverability, content structure, internal linking, and technical SEO. Real audits on live documentation have surfaced findings like missing breadcrumb navigation across all sampled pages, meta descriptions absent or client-side injected, canonical URLs not confirmed, install commands lacking version pinning, and absent central troubleshooting indexes.
These are exactly the gaps that prevent developer documentation from being cited by AI search engines, even when the documentation itself is technically accurate and well-written.
4. SEO Team Skills
The SEO pack ships 26 open-source skills specifically for SEO teams. The skill set covers technical audits, schema validation, GEO and AI search readiness scoring, local SEO signals, backlink analysis, and content scoring.
Notable skills in this pack include:
- orphan pages audit: Finds all blog and content pages with zero incoming internal links, clusters them by topic, and generates three linking suggestions per orphan with anchor text and placement guidance.
- No-outlinks audit: Identifies dead-end pages with no outgoing internal links, preventing link equity from circulating through your content graph.
- llms.txt checker: Audits whether your domain has an llms.txt file, whether it is structured correctly, and gives prioritized fixes if it is missing or incomplete.
- Growth report: Generates a 90-day SEO performance report pulling live data from your domain, covering traffic trends, keyword rankings, top content pages, and competitive positioning.
The SEO pack integrates with Ahrefs via API for keyword data enrichment. Most skills in this pack work without any external API keys, and the Ahrefs integration is additive rather than required.
5. Web Design Skills
The web design pack covers UI design direction, landing page audits, site architecture planning, and design guidelines reviews. It is built for design teams that want AI-assisted feedback on visual and structural decisions, not just content.
How to Install Claude Skills in Three Steps
Setup takes under ten minutes and requires no configuration files, no environment variables for most skills, and no prior familiarity with Claude Code.
Step 1: Clone the repository
git clone https://github.com/Infrasity-Labs/dev-gtm-claude-skills.git
Use a shallow clone if you want to keep the footprint small.
Step 2: Enter the repository and open Claude Code
cd dev-gtm-claude-skills
claude
Claude Code reads the repository structure and loads all available skills automatically when it starts inside the directory.
Step 3: Run your first skill
dev-gtm docs-audit https://your-docs-site.com
Replace docs-audit with any skill name and the URL with your target. The skill handles the rest, crawling the URL, running its checks, and producing a structured report with a prioritized fix list.
Skills also upload directly into Claude.ai if you prefer the browser interface. In that case, upload the relevant SKILL.md file at the start of your session.
Real Audit Results: What These Skills Surface
Looking at published audit results from live documentation sites gives a clear picture of what these skills catch that traditional tools miss.
Documentation site with 83/100 overall score: The audit surfaced missing breadcrumb navigation on all sampled pages, meta descriptions not confirmed on any sampled page (likely client-side injected and therefore invisible to crawlers), canonical URLs absent, install commands without version pinning, and no central troubleshooting index. None of these issues would appear in a standard Lighthouse or Screaming Frog report.
API documentation platform with 22 of 84 endpoints passing: The audit found that 25 Twitter Read endpoints were missing all descriptions and error codes, that 15 or more OpenAI Chat body parameters were undocumented (including temperature, max_tokens, and tools), that only a 200 response code was documented across more than nine endpoints, and that a deprecated endpoint remained live in production with no deprecation notice visible on the page.
Deployment platform SEO growth report: The 90-day analysis showed an 18.1% traffic decline despite holding the number one ranking position for core keywords. The report identified keyword count collapsing by 47,801 terms as the single largest SEO vulnerability, flagged 2,054 keywords sitting in positions 4 to 10 as the highest-priority recovery pipeline, and noted that the domain was still the most resilient in its competitive set (competing platforms were down 29 to 38 percent over the same period).
These are the kinds of insights that take hours to surface manually and that most automated SEO tools do not check at all.
Who These Skills Are Built For
Claude skills in this repository are optimized for teams building and marketing technical products, specifically:
DevTool startups building CLI tools, APIs, SDKs, or developer platforms. If your product is adopted through documentation, these skills audit that documentation against the signals that determine whether a developer finds it via AI search.
AI agent platforms with LLM applications, autonomous agent frameworks, or AI workflows. These products are evaluated by AI search engines that understand the space. Documentation quality is a direct competitive differentiator.
Observability and infrastructure companies whose buyers are engineers conducting technical evaluations. When a site reliability engineer asks Perplexity to compare APM tools, the tools with the strongest AI citability signals appear in the answer. The others do not.
B2B SaaS products with technical buyers who have long evaluation cycles. Developer documentation that reads well to an LLM reads well to an engineer. The two audiences overlap more than most marketing teams realize.
For teams that want audits run and fixes implemented rather than a CLI tool to run themselves, the developer marketing agency for DevTool startups operates the same workflows at a managed service level, measured on product adoption metrics like API calls and SDK installs rather than vanity traffic numbers.
Teams specifically focused on AI search visibility can work with GEO and AEO optimization services that optimize content for citation across ChatGPT, Claude, Perplexity, and Gemini.
GEO Best Practices for Developer Documentation
Getting your documentation to appear in AI-generated answers requires attention to a specific set of structural and semantic signals. Based on what the Claude skills audits surface most consistently, here are the highest-impact changes you can make:
Add or fix your llms.txt file. This file tells AI systems how to read and cite your documentation. It should be present at the root of your domain, structured correctly, and include pointers to your key documentation sections.
Ensure breadcrumb navigation is server-side rendered. Client-side injected breadcrumbs are invisible to crawlers and to AI systems parsing your HTML. Navigation context helps LLMs understand the hierarchy of your documentation.
Add version pinning to all install commands. Install commands without version pinning are an AI citability risk. LLMs parsing your quickstart guides flag ambiguous install instructions as lower-confidence sources.
Build a central troubleshooting index. Scattered troubleshooting pages without a navigable index prevent AI systems from establishing the authority and completeness of your documentation. A single, well-structured troubleshooting hub changes this.
Increase internal link density. Orphan pages with no incoming internal links are effectively invisible to both search engines and AI systems. Every documentation page should have at least one contextually relevant incoming link from another page on the same domain.
Confirm canonical URLs server-side. Canonical tags set via JavaScript are unreliable. Server-side rendered canonicals give both search engines and AI crawlers a clear, authoritative signal about which version of a page should be cited.
Top comments (0)