EN
Keep me posted
All agents

An agent at work · AI coach

From a rough questionto a prompt that works.

The AI coach is one of the agents in Zaela. It turns whatever you send into a sharper prompt, gives concrete tips to get more out of Claude and explains the principle, so you can do it yourself next time. Before it says anything about a feature, it checks the current documentation.

The film

AI coach

Made in code with Zaela. The voice is synthetic.1:03

The method

This is how the AI coach works.

  1. 01A better prompt

    Every input becomes a better prompt.

    Whatever you send, it rewrites into a sharp prompt: the goal, the context, the format you want, an example or a role. It does this unasked, as soon as it sees that something can be sharper, and says briefly why the new version works better.

  2. 02The goal

    What are you really trying to do?

    It looks for the goal behind the question and asks further when it is truly vague. When the question is about a brand or a company, it reads Zaela’s shared memory and also searches online, to understand where that company stands today. It keeps clear what is fact and what is assumption.

  3. 03Current docs

    Today’s documentation, not yesterday’s memory.

    Claude changes every month: new models, new features, renamed buttons, other limits. So before it says anything about a feature, a model, a limit or a button, it looks it up live in the official documentation and gives you the source and the date.

  4. 04Your environment

    A tip for the place you work.

    A slash command, a button or a shortcut can differ between Cowork, Claude Desktop, Claude Code, the browser and the API. It first checks which environment you are in, asks when it does not know, and always says where a tip applies.

  5. 05Tips and the way

    The right tip, and the right colleague.

    It adds tips that fit the task: choosing a model, giving a role and context, working in phases on a big task, keeping decisions in md files so nothing is lost when the context runs out. And it points you to what already exists in Zaela for the job: an agent, a skill, a tool or a memory space.

  6. 06The principle

    So you can do it yourself next time.

    It explains why the better prompt works and leaves you with one concrete tip. It keeps things simple and encouraging, and assumes little technical knowledge. And it reminds you that these are today’s best practices: tomorrow they may be different.

This page is based on the agent's instructions and knowledge in Zaela. The drawings are illustrations.

What it pays attention to

What it pays attention to.

  1. 01

    Proactive

    It improves your prompt even when you did not ask, and shows the better version straight away.

  2. 02

    Source and date

    Every statement about a Claude feature comes with the official page and the date. It says openly that this changes often.

  3. 03

    Never made up

    No invented feature, model name, button or number. When in doubt, it looks it up or says so.

  4. 04

    Where you work

    A tip that only works in one environment is never presented as if it works everywhere.

  5. 05

    Fact or assumption

    It always makes clear what comes from a source or data, and what is an assumption.

  6. 06

    Simple first

    It starts from the basics when that helps: what an AI model is, why it can be confidently wrong, what MCP does.

Teamwork

Who it works with.

What it never does.

  • Make the deliverable itself. It points you to the right agent, skill or tool.
  • Rely on its own memory for a feature, a price, a limit or a button.
  • Invent a feature, a model name or a number.
  • Present a tip as if it works in every environment.
  • Present an assumption as a fact.

What you get

What you get.

Curious what it could do for your organisation?

Agents at work

All agents

Stay informed

Curious what Zaela could become for you?

Leave your details and we’ll get in touch to tell you more about Zaela.

We use your details only to contact you about Zaela.