Start

Directory · read every fortnight

The agent skills a search and marketing team would actually install

Every skill on skills.sh, read every fortnight and filed under the categories this audience works in. Their install counts and their security audits, our reading of what each one is for, and the trend since we started counting — which a live leaderboard cannot show you.

Source

skills.sh

Directory, counts and audits. Linked on every row.

Skills tracked

155

15 filed under a category this audience uses

Last sync

September 3, 2026

Read on September 3, 2026

Last cycle

153

153 updated · 0 withdrawn

Filed by what it does for you

The directory has no category for search work, so these are ours: every skill is read once and filed, and the chip on the row says so.

All155Technical SEO44GEO / AI search15Content27Analytics18Paid3Agent workflows48

Showing 1–15 of 15

Install counts and audits are skills.sh's, shown as they were at the last sync. The trend is ours: it exists because we keep the previous readings, which a live leaderboard does not.

Audit: what independent security auditors found in the skill's own files, reported through skills.sh. A skill is instructions your agent will follow and scripts it may run, so this is worth as much as the install count. Pass means every auditor that looked found nothing; warn and fail mean at least one did — hover a verdict to see who said what. None yet means nobody has audited it: the directory generates an audit after a skill's first install, so it is a fact about the skill's age, not about its safety.

Our brief: the languages our own reading of the skill is published in. Every skill page is served in all five languages — these chips are about the brief inside it, which is written one language at a time and reviewed before it appears.

Questions people ask

What is an agent skill?

A folder of instructions, and sometimes scripts, that an AI coding agent loads to do a specific job — write an llms.txt, audit redirects, pull a Search Console report. You install one with a single command and the agent follows it from then on.

Where does this list come from?

From skills.sh, through its documented API, read every fortnight. Install counts and security audits are theirs, shown as they were at the last sync and linked back on every row. Which shelf a skill sits on is ours.

Why are these categories not the ones on skills.sh?

Because their API has no category filter at all, and the topics on their pages are a developer's shelves — React, databases, testing. A marketer looking for something that makes their pages retrievable is not helped by either, so we read each skill once and file it under the work it does.

What does the trend column mean?

The change in install count since the first time we read that skill. The directory publishes today's number; we keep every number we have taken, so the shape of the climb exists here and nowhere else. A skill read once says so rather than showing a change of zero.

Is a skill with no audit unsafe?

No — it is unaudited, which is a different fact. The directory generates an audit after a skill's first install, so an empty cell usually means the skill is new or rarely installed. Hovering a verdict names every auditor and what each one said.

Do you host the skills?

No. A skill belongs to its author under the licence in their repository. We show a name, a shape, a count and a link; installing happens on the directory or in the repository, never here.

Asked out loud

spoken, not typed

The same questions in the words somebody uses speaking to an assistant rather than typing into a box. Each one carries the situation it came from, because the situation is what makes it a different question.

What skills should I install for SEO work?

Start from the Technical SEO shelf here, sorted by installs: a skill thousands of people run is one whose failure modes are known. Check the audit column before you install anything that will run scripts, and prefer one with more than one auditor behind its verdict.

setting up an agent for the first time
Is there a skill that writes llms.txt?

Yes, and it is one of the most installed things on the GEO shelf. The skill reads a sitemap, ranks the pages worth naming to an assistant and keeps the file in step with the build — which is the part that usually rots when somebody writes one by hand.

after reading about llms.txt
Which of these is actually growing?

The trend column answers that, and it is the only column here a live leaderboard cannot show you: it compares today's count with what we recorded a fortnight ago. A skill climbing steeply is a job enough teams have that somebody bothered to automate it.

deciding what to learn next
Can I trust a skill from a repository I have never heard of?

Read the audit column first: it reports what independent security vendors found in the skill's own files. Then open the skill's page — the file list shows how big the thing is and what it carries, and the link goes to the repository where you can read it yourself before running it.

before installing something unfamiliar

The words on this page

Agent skill
Instructions for one job, filed where an agent can find them and loaded only when that job comes up. This is what the directory lists.
Skill
The word alone means three things — an agent's instructions, a voice-assistant app, or a plain capability. Which one is meant changes the answer entirely.
Model Context Protocol
The open standard for connecting an assistant to outside systems. A server gives an agent reach; a skill gives it know-how.

Provenance and limits

  • Data from skills.sh through its documented API. Every row links back to the skill's page there.
  • Skills belong to their authors under the licences in their repositories. We show a name, a shape, a count and a link — never their files.
  • Categories, trends and the reading of what each skill is for are ours, and are marked as ours.
  • Nothing published here is in Russian. Rows whose text is Russian are refused before they are stored, and the count of refusals is kept with the sync.