A prompt tells an AI agent what you want this once. A skill tells it how the work is done where you are, every time that kind of work comes up. Skill Publisher is where we write skills to order.
An agent skill is a folder containing a SKILL.md file: Markdown instructions with a short block of metadata saying what the skill is for and when an agent should use it, plus any reference files it needs. It follows the Agent Skills specification, which Anthropic published in December 2025 and released as an open standard. An agent reads the description, decides whether the skill is relevant, and loads the rest only when it is.
If a question came out the same either way, you will see that.
Why a skill is worth more than a prompt
A prompt lives in one person's tool. A skill is a folder of plain Markdown in your repository, which your team can read, edit and version like any other file. Around forty agent products read the format, including Claude Code, ChatGPT and Codex, GitHub Copilot, Cursor and Gemini CLI. Switch editor, model or vendor and the skill comes with you.
The format is simple. The craft is deciding what goes in it. Ask a model to write its own skill and it writes what it already knows, which is the one thing the agent did not need telling. A skill is worth loading when it carries what the model lacks: your conventions, your exceptions, the way the work is done where you are.
How an order runs
- You send a brief: what your agents should get better at, and three to ten real questions the skill has to handle
- Our research team analyses the field, and anything you send: documents, handbooks, recordings by link
- We write the skill, containing only what your agents do not already know. Every claim in it is checked against the source it came from
- We run your questions with the skill loaded and without it, and hand you the folder, both sets of answers and the install path for each tool you use
Turnaround is ten working days from the point we accept the brief. Skill packs start from £39 + VAT, and current prices are on the Skill Publisher pricing page.
Why we test with and without
Writing a skill looks easy, because it is a Markdown file. That is exactly why most of them do not work. They read well, they load, and the agent behaves the same as it did before. Until you run the same question with the skill and without it, there is nothing to compare, so nobody finds out.
That is why the test questions are yours, not ours, and why you get both answers side by side. If a question came out the same either way, you will see that. We would sooner show you where a skill made no difference than ask you to take our word that it helped.
Standards for coding agents
Where the gap is a published standard such as ISO 25010 or OWASP, or your own engineering rules, the sister route is Code Standards, which provides standards packs for coding agents in the same format.
Both routes are described on our Agentic Skills page. If you would like to talk through what your agents are missing first, get in touch.
