⌘K
DARK
MENU
DARK

Free library

Skills-md: free SKILL.md files for AI agents, Claude & Codex

Skills-md is a free library of SKILL.md files: ready-made instructions that teach Claude, OpenAI Codex and other AI agents to do professional work the way a senior practitioner would. Every skill is one Markdown file you can read, copy or download. Marketing is live now; productivity, product management, video, data and finance skills follow, in that order.

By Afeez Adebayo, Marketing, SEO & Growth Expert with 10+ years of experience.

Created 9 October 2026. Updated 9 October 2026. Free to use under CC BY 4.0.

Follow me onXLinkedIn

Definition

What is a SKILL.md file?

A SKILL.md file is a Markdown document with a short YAML header (a name and a description) followed by instructions. An agent reads the description to decide when the skill applies, then loads the full instructions only when a task needs them.

The library today

250skills published

250 of 299 planned skills across marketing and productivity (84%), every one at least 1,500 words.

21 / 23subcategories live
2 / 6categories live: marketing and productivity
4categories next: product management, video, data and finance

Marketing

Productivity

All 250 skills in one folder (.zip)Install-ready: folder per skill (.zip)

Skill categories

Each category is split into subcategories, and each subcategory into 10 to 15 skills you can install on their own or as a set.

03

Product Management

Planned, next

PRDs, discovery interviews, prioritisation, roadmaps, launch plans and product analytics reviews.

04

Video

Planned, in this order

Script writing, editing, video production, voice-overs in different tones, genders and languages, sound and visual effects, short films and 15 to 60 second video ads.

05

Data

Planned, in this order

SQL reviews, analysis plans, experiment readouts, dashboards and data-quality checks for data scientists and analysts.

06

Finance

Planned, in this order

Reconciliations, month-end close, variance analysis, budgets, forecasts and management reporting for accountants and financial analysts.

On this page

  1. What is a SKILL.md file?
  2. How SKILL.md files work in Claude and Codex
  3. How to install a skill in Claude Code, Codex and Claude.ai
  4. How to choose which skills to install
  5. How every skill in this library is built
  6. Who writes and maintains these skills
  7. Frequently asked questions
  8. Sources

What is a SKILL.md file?

A SKILL.md file is a plain Markdown document with a short YAML header. The header holds two required fields: a name, which must match the folder the file sits in, and a description that says what the skill does and when to use it. Everything below the header is instructions for the agent.

The format comes from the open Agent Skills specification published at agentskills.io. Claude reads it, and OpenAI documents the same structure for Codex, so one file works in both. Optional fields such as license and metadata let an author add credit and versioning without changing how the skill behaves.

A skill is not a prompt you paste into a chat. It is closer to a standard operating procedure the agent keeps on a shelf: the inputs to ask for, the steps to follow in order, the rules that decide between options, and the shape of the finished output. Because the agent only reads it when a task calls for it, you can install many skills without filling its context.

How SKILL.md files work in Claude and Codex

Agents load skills in three stages, which the specification calls progressive disclosure. At the start of a session the agent reads only each skill’s name and description, roughly a hundred tokens per skill. When your request matches a description, it loads the full SKILL.md body. If the body points to extra files, such as a reference table or a script, it opens those only when a step needs them.

  1. 01You askA plain request, for example "Audit our sitemaps before the migration."
  2. 02Agent matchesIt reads every installed skill's name and description and picks xml-sitemaps.
  3. 03Skill loadsOnly now does it read the full SKILL.md: inputs, workflow, rules, template.
  4. 04Structured outputYou get the audit in the skill's output format, with sources to check.
How an agent uses a SKILL.md file. Based on the progressive disclosure model in the Agent Skills specification (agentskills.io), October 2026.

This is why the description matters more than any other line in the file. It is the only part the agent sees before deciding, so it has to name the task and the words people use when they ask for it. Every description in this library lists those trigger phrases, for example “why did traffic drop” for the technical SEO audit.

The specification recommends keeping the main file under 500 lines and about 5,000 tokens. The skills here stay inside that limit, so they load quickly and leave room for your own files and data.

How to install a skill in Claude Code, Codex and Claude.ai

Every skill lives in its own folder, and the folder name must match the name field in the file. Download a single skill as a .md file, rename it SKILL.md and put it in a folder with the skill’s name, or download a whole set as a zip that already has the right structure.

Claude CodePersonal or project skills
# all projects
~/.claude/skills/xml-sitemaps/SKILL.md

# one project
.claude/skills/xml-sitemaps/SKILL.md

Start a new Claude Code session and ask which skills it can use.

OpenAI CodexUser or repository skills
# all repositories
~/.agents/skills/xml-sitemaps/SKILL.md

# one repository
.agents/skills/xml-sitemaps/SKILL.md

Codex reads .agents/skills folders from the working directory up to the repo root.

Claude.aiWeb and desktop apps
1. Put SKILL.md in a folder named xml-sitemaps
2. Zip the folder
3. Customize > Skills > Upload a skill

Code execution must be enabled for skills on your account.

Personal folders make a skill available in every project. Project folders keep it with one codebase or client, which suits agencies and teams who want the same instructions shared through version control.

How to choose which skills to install

Install the set that matches the work in front of you rather than the whole library. Agents choose between skills by reading descriptions, and a long list of similar descriptions makes a wrong match more likely.

  • Start with one subcategory, for example Technical SEO, and use it on a real task for a week.
  • Keep the audit or overview skill installed; it routes to the specialist skills when a finding needs depth.
  • Remove skills you never trigger. A smaller set is easier for the agent to choose from and for you to maintain.
  • If you adapt a skill for your company, keep the name the same and add your own terms to the description.

How every skill in this library is built

Each file follows the same structure, so you always know where to look:

  1. When to use it, and when not to, so the agent doesn’t apply it to the wrong job.
  2. The inputs to ask for, in one message, so you aren’t asked the same question twice.
  3. A numbered workflow in the order an experienced practitioner would follow.
  4. Decision rules with thresholds, such as Core Web Vitals limits or which redirect to use.
  5. An output template, so results come back in a format a team can act on.
  6. Common mistakes, taken from real projects.
  7. Links to the official documentation each rule depends on.

Facts that change, such as which rich results Google still shows or the names of AI crawlers, carry a date and a source link, and the skill tells the agent to re-check the source before relying on them. Version 1.0.0 of the Technical SEO set was checked against Google Search Central, web.dev, the Agent Skills specification and OpenAI’s Codex documentation in October 2026.

Who writes and maintains these skills

Frequently asked questions

Yes. Every skill is released under the Creative Commons Attribution 4.0 licence. You can copy, share and adapt it, including for commercial work, as long as you credit Afeez Adebayo and link to the page you got it from. The notice at the end of each file already does this; keep it in any copy.

They work with any agent that reads the Agent Skills format. That includes Claude Code, the Claude apps and OpenAI Codex, which loads skills from .agents/skills folders. Other tools that support the standard can use the same files without changes.

A prompt is something you type each time. A skill is installed once and loaded by the agent when a task matches its description. It also carries more than a prompt usually does: the inputs to ask for, a workflow, decision rules, an output template and sources.

Only a little. Agents read each skill’s name and description at the start, about a hundred tokens each, and load the full file only when it is needed. The bigger risk with too many skills is a wrong match, so install the sets you use.

Yes. Add your tools, thresholds and house style, and keep the credit notice at the end of the file. If you publish your adapted version, link back to the original skill page so readers can find updates.

Each file has a version number and an updated date in its header. I review a set when Google, OpenAI, Anthropic or another platform changes the documentation it relies on, and note the change on the skill’s page.

Marketing is being completed one subcategory at a time: Technical SEO, Content and On-Page SEO and AEO & GEO are published. Then come, in this order: productivity, product management, video (script writing, editing, voice-overs, effects and short video ads), data, and finance.

Sources

All sources checked on 9 October 2026.