Data

Official AI agent skills statistics

A market report on official AI agent skills: which companies publish them, what their skills teach agents to do, and which software segments already have real coverage.

Data collected July 19, 2026. Article updated July 21, 2026.

What skills are

AI agent skills are reusable instruction packages that teach an agent how to complete a specific type of work. A skill can include a trigger, step-by-step workflow, examples, references, scripts, tool rules, and safety checks.

In practice, a skill turns product knowledge into agent-ready procedure. It can teach an agent how to call an API, configure auth, review a pull request, deploy an app, create a dashboard, run a security check, or follow a company workflow.

The open Agent Skills format was originally developed by Anthropic and published as an open standard. The Agent Skills overview defines the basic shape: a folder with a required SKILL.md file and optional scripts, references, assets, or other resources. The specification keeps the format small enough for companies and agent tools to adopt, with development continuing in the open GitHub repo.

Official skills are a new SaaS surface

Official skills are published by the company, project, or product team that owns the underlying software. They are becoming a new SaaS surface because they let companies ship product knowledge directly to AI agents.

A SaaS company already has docs, SDKs, dashboards, APIs, support articles, and templates. An official skill packages the practical path through those materials: which API to use, which setup steps matter, which defaults are safe, and which checks should happen before production use.

This changes how software can be adopted. A user can ask an agent to add analytics, payments, observability, authentication, or deployment support, and the agent can follow vendor-owned instructions instead of guessing from scattered examples.

What companies usually turn into skills

The most common official skills teach agents how to work inside software systems with less guesswork. API and SDK usage is a major pattern. Companies also publish skills for code review, deployment, analytics, observability, setup, security, research, documents, and product workflows.

The table below groups public skill names and descriptions into recurring themes. A skill can appear in more than one theme, because product work often overlaps. For example, a Supabase RLS skill is both infrastructure and security, while a Stripe Checkout skill is both API work and payments.

Code and API work lead because official skills are closest to developer tools today. The next large group is operational: deployment, data, debugging, analytics, and setup. Companies use skills to package repeatable product work with instructions, examples, checks, and product-specific procedures.

Which segments are covered

Official skills are strongest in developer infrastructure, AI tooling, cloud deployment, data access, analytics, and product engineering. Business-team coverage is emerging through design, content, productivity, payments, commerce, and customer-facing tools.

What the leading companies offer

Companies publish official skill repos or skill libraries so builders can install vendor-owned instructions for agents.

Microsoft has the broadest company-owned surface, with many skills for Azure and business platform operations. Anthropic publishes skills around Claude workflows, document formats, frontend implementation, legal and finance workflows, and MCP. OpenAI publishes skills connected to Codex, repo work, security review, browser testing, Figma, Linear, and app-building workflows.

GitHub is strong in software delivery workflows through Copilot skills. Vercel connects skills to deployment, web design, the AI SDK, and agent browser work. Firecrawl gives agents a complete search and extraction toolkit. Supabase, Firebase, Cloudflare, and Stripe show the next layer: official skills for infrastructure, auth, storage, observability, and payments.

Smaller company skill libraries can still be useful. Metabase has 11 skills focused on embedded analytics and semantic checks. Mixpanel has 26 skills for dashboards, tracking, experiments, feature flags, and lexicon management. Canva has a compact set of design workflow skills for bulk creation, resizing, brand checks, feedback, and presentations.

What this says about the market

Official skills are moving fastest where a product has developer adoption, public docs, APIs, SDKs, CLIs, or dashboards. These products already need precise workflows, and agents benefit from vendor-owned instructions.

The current official-skills market is centered on developers, engineers, and the agents they use for software work. The strongest categories are code review, API and SDK usage, deployment, analytics instrumentation, debugging, observability, and infrastructure setup.

Skills for everyday business users are at an earlier stage. Design, content, productivity, payments, commerce, and customer support have less depth than engineering workflows. The likely next wave is skill libraries that help operators, marketers, analysts, founders, and support teams use complex SaaS tools through agents.

Official skills also change product distribution. A company can publish an official skill so agents learn the preferred API path, common setup steps, safe defaults, and troubleshooting flow. That makes official skills part of documentation, support, onboarding, and developer relations.

How builders should use this data

Builders should start with official skills when the task involves a vendor product: authentication, payments, cloud setup, analytics, observability, design, or API work. Vendor-owned skills usually encode current product assumptions better than random examples.

The useful workflow is simple: search the Skillscout official skills directory, open the source, read the SKILL.md, check scripts and permissions, then install the skill into the agent environment you use.

What this report counts

An owner is a company, organization, lab account, or project that publishes official skills. A repo is a public source repository that contains one or more skills. A skill is a reusable agent instruction package centered around a product, workflow, API, tool, or domain task.

Skill count measures how much surface area a company exposes to agents. GitHub stars measure repository visibility. Install counts measure demand on directories and marketplaces that report installs. The rankings are separated because each signal answers a different question.

Data caveats

Install counts come from directories and marketplaces that expose them. GitHub stars come from repositories, so a large repo can lift a company even when the skill set is small. Some source directories split brands across lab accounts, product accounts, and main organizations. Future editions can separate strict verified-company rankings from broader official directory rankings.

The market direction is clear: official AI agent skills are concentrated around developer tools and infrastructure today, with analytics, design, productivity, payments, and commerce expanding behind them.

Statistics

The statistics below come from the Skillscout official skills directory, which tracks official skills across SKILLS.sh, officialskills.sh, mcpservers.org, Cursor Marketplace, and GitHub. The current totals are:

Official owners publishing the most skills

Microsoft has the largest official-skill footprint, followed by Anthropic and OpenAI. The top rows show two patterns: broad platform companies publish many small skills across products, while AI-native companies publish workflow libraries for agent building, coding, files, research, and MCP.

Official owners with the strongest install signals

Install counts highlight demand on discovery and marketplace surfaces. Microsoft leads because its Azure skills have very high install counts. Vercel Labs, Anthropic, GitHub, Firecrawl, Firebase, Remotion, Expo, and Supabase also show strong marketplace demand.

Official owners with the strongest GitHub visibility

GitHub stars favor companies with popular open-source repos. Stars are useful for ecosystem visibility and weaker as a direct skill-usage metric. The leading rows show where skills are attached to widely watched developer ecosystems.