An AI agent from any provider can work in the PromptEye style. It does not have to guess from screenshots: it gets the same rules, tokens and files as our team, in the formats models read best.
Three ways
Ready prompt
Paste it at the start of a conversation or into the project instructions.
Use the PromptEye identity system. First read https://brand.prompteye.com/llms.txt, then the documents it points to for what you are building: start with “Start” and “Formats”, then the chapter of the block you use.
Take colours, fonts, logo and layout only from this system. In HTML load styles.css and the scripts from https://brand.prompteye.com/ds/. In Word, PowerPoint and Google Docs use the colour tokens, the Geist and Inter fonts and the SVG logo, without imitating the motif.
Go through the checklist before you deliver. If something is not in the system, ask instead of inventing it.
By tool
| Tool | What to do |
|---|---|
| Claude | Upload the ZIP as a Skill or add the llms.txt address to the project instructions. In Claude Code unzip the prompteye-brand folder into your skills directory. |
| ChatGPT | Add AGENTS.md and the docs/ folder as project files. With web search on, the llms.txt address in the prompt is enough. |
| Gemini | Add AGENTS.md and docs/ as knowledge in a Gem, or give it the llms.txt address. |
| Codex, Cursor, Copilot, Windsurf | Add the unzipped kit to the repository. These tools read AGENTS.md automatically or on request. |
| DeepSeek, Mistral i modele bez internetu | Attach llms-full.txt (PL or EN) or paste it into the chat. |
Addresses for agents
All addresses are public and stable. Every page’s content is in HTML without JavaScript, so any page-fetching tool can read it.
| Address | Contents |
|---|---|
/llms.txt | index: rules in brief and links to every document |
/llms-full.txt | all documentation in one file, Polish (source) |
/en/llms-full.txt | all documentation in one file, English |
/en/<chapter>.md | every page in Markdown, e.g. /en/color.md |
/ds/AGENTS.md | instructions for agents (same as SKILL.md) |
/ds/tokens.json | tokens in DTCG format |
/ds/styles.css | styles: fonts, tokens, components |
/ds/js/pe-motyw.js | motif script (plus pe-light, pe-pl, pe-raster) |
/ds/assets/logo/svg/ | logo in SVG |
/downloads/prompteye-brand-kit.zip | kit: SKILL.md, AGENTS.md, docs, tokens, CSS, JS, logo |
Language
The source documentation is in Polish. English translations are added chapter by chapter; the English address of an untranslated chapter returns the Polish text with a note. Token, class and attribute names are Polish words (--pe-kobalt, data-pe-hero) and are used exactly as written. The agent writes the material in the user’s language.
What an agent cannot do
- The motif and grain gradients are drawn by JavaScript in a browser. In Word, PowerPoint, Google Slides and Canva an agent cannot recreate them and should not imitate them.
- Items “to be developed with the website” are not part of the system yet. The agent should ask about them.
- The agent writes marketing copy according to the Idea and messaging chapter, using only confirmed numbers.