Skip to content
Crawl = AI promptsWalk = AI skillsRun = AI agents
kyee.ai
Skill
Free

AI Corpus

The skill explains in plain terms what each major AI model was built and trained on, and what that makes it good or bad at. You get a simple comparison, so you can pick the right model for your kind of work.

AI & Automation

Plain-text .skill.md file · Setup guide included · Runs in your own AI account · No coding needed.

Supports:
ChatGPTClaudeGeminiGrokMistral

A LOOK INSIDE YOUR SKILL

A look inside your skill

ai-corpus.skill.md · Partial preview

---name: ai-corpusdescription: >-  Maps what data an AI lab is entitled to train on and how that differs from  other labs. Takes a model or lab name, researches the internet, and builds a  rights map: the data the lab can use because of who owns, funds and partners  with it (first-party data, licences, partner agreements, user-consent terms),  the limits on each right, and how much is confirmed as actually used in  training. Also covers the base corpus every large model shares. Every line  carries an evidence grade. Interviews the user with one or two simple  questions and explains the result in plain language. Use when the  user says "ai corpus", "what was [model] trained on", "training data for  [model]", "what data is [lab] allowed to use", "what does [model] know that  others don't", "which model knows [domain] best", "corpus of [lab]", "what  data does [lab] have", or wants to pick a model for a task based on what it  has seen.---# AI Corpus## Overview**What it does.** Tells you what an AI model most likely learned from, and what it can learn from that other models cannot. Every AI company has a different pool of data it is allowed to use, depending on who owns it, who funds it, and who it has deals with. This skill maps that pool for the model you name and says how sure we can be about each part.**When to use it.** When you are choosing a model for a job and want to know which one has seen the most relevant material, or when you want to understand why one model seems to know a subject better than another.**How it works.**1. **It asks** which model or company, and whether there is a subject you care about most (code, law, medicine, another language).2. **It researches** who is behind the company and what data each owner and partner holds, what the company has published, what its web crawlers collect, the deals it has signed, what its terms say about using customer data, and any lawsuits.3. **For each data source it answers three separate questions:** Is the company allowed to use it? What limits apply? Is there evidence it really is used?4. **A checker** confirms each claim against the actual terms, announcement, or filing.5. **It reports** in plain language.**How sure is each line?**- **Confirmed**: the company or its partner said so.- **Reported**: others say so; the company has not confirmed.- **Likely / Possible / A long shot**: a careful guess based on who owns the company and what it has access to.**What you get back.** A short answer in the chat, at most five points on what makes this model different, and a saved report covering what every model learned from, what this company can use that others cannot, what they say they do not use, deals they have made, and what nobody outside the company knows.**Good to know.** No company publishes a full list of its training data, so much of this is informed guessing, always labelled as such. It is not legal advice. Being allowed to use data is not the same as using it, and a model looking something up live is not the same as having learned it; the report keeps these apart. Everything is researched fresh on each run because this changes fast.## Examples## Dependencies## How to talk to the user## 1. Input## 2. Three evidence grades## 3. The rights map## 4. Research## 5. Check## 6. Output## The short version## What this company can use that others cannot## Who is behind it## What everyone's model learned from## What they say they do not use## What it looks up live instead of learning## Practice examples made by people or by other AI## Deals they have made## What is being argued about## What this means for you## What nobody outside the company knows## Sources## How this was checked## Rules

23 protected steps and quality checks in the complete skill

WHY USE A SKILL INSTEAD OF A QUICK CHAT?

The difference between a quick chat and a tested skill.

Everyday prompting
  • A quick, general answer
  • Confident guesses
  • No clear next step
With a KYEE skillAI Corpus
  1. A structured, step-by-step process
  2. Gaps and uncertainty made clear
  3. A result you can act on

Verification system

We've built in the know-how.

  1. The know-how

    30+ years in tech, business experience, and deep AI expertise.

  2. The method

    Clear steps and quality checks, you don't have to guess how to ask AI.

  3. The refinement

    We check results against clear standards and refine the answer.

KYEE’s internal method, not an independent certification. How skills are checked

Details

Skill specifications

File format
Plain-text .skill.md file (Markdown with a short header) containing the steps, rules and quality checks
How it runs
Paste into your AI chat box — no plugins, API keys or IT setup
Works in
ChatGPT, Claude, Gemini, Grok and Mistral
Needs
Web search turned on in your AI app, to check current sources
Your data
Runs inside your own AI account. Review your AI provider’s data terms before sharing sensitive business information.

Questions you might have

What is a KYEE AI skill?

A plain-text instruction file (.skill.md) that guides your AI app through a business task step by step, with built-in quality checks. You paste it into your own AI chat.

Which AI apps does it work with?

It is designed for chat apps that accept pasted instructions, such as ChatGPT, Claude, Gemini, Grok and Mistral. Some features, like web browsing, depend on your app and plan.

Do I need an AI subscription?

You use your own AI app, free or paid. An AI subscription is not included. Some tasks work better on paid plans.

Do I need technical skills?

No. Each skill includes a setup guide written for busy professionals. You copy the skill, paste it into your AI chat and follow the steps.

What is included?

The skill file and a setup guide.

Can I try KYEE skills for free?

Yes. Some skills are free — you only tell us where to send them, so you can see how they work before buying.

How does payment work?

Paid skills and Workflows are planned as one-time purchases in USD. Checkout is not live yet, so no payment can be taken on this site today.

Does KYEE see my chats or business data?

The skill runs inside your own AI account, so your chats stay with your AI provider. Review your provider’s data terms before sharing sensitive information.

What is the B.S.C. Verification System?

KYEE’s own internal method for building and checking skills: know-how, clear instructions and refinement. It is not an independent certification and does not guarantee error-free AI results.

Should I check the AI’s results?

Yes. AI can make mistakes. Review sources, figures and recommendations before acting, and ask a qualified professional for legal, financial or tax decisions.

Want to know how skills are checked? About the Verification System