kubical-ai/ai-prompt-files

AI Prompt Files Creator

Scaffolds AI editor context files that give coding AI assistants deep, accurate knowledge of a project.

Load assets/prompt-template.md for the recommended content structure template.

When to Use This Skill

• Creating .cursorrules, CLAUDE.md, or .github/copilot-instructions.md for a docs project (item #12)
• Setting up AI coding tool context for any project
• Updating outdated AI context files

---

File Targets by Tool

| Tool | File | Location | |------|------|----------| | Cursor | .cursorrules | Project root | | Claude Code | CLAUDE.md | Project root or .claude/ | | GitHub Copilot | copilot-instructions.md | .github/ | | Cline/Roo | .clinerules | Project root | | Windsurf | .windsurfrules | Project root |

Content is identical across files — the filename determines which tool reads it. Always create at least CLAUDE.md + .cursorrules for maximum coverage.

---

Workflow

Step 1: Understand the project deeply

Read the following to gather context:

1. README.md — overall project description
2. package.json — dependencies, scripts, framework version
3. Directory structure — what's in src/, docs/, scripts/
4. A sample of content files — what kind of docs are these? API reference? Guides? Tutorials?
5. Key config files — astro.config.mjs, next.config.js, docusaurus.config.js
6. Any existing API documentation — endpoint patterns, authentication approach

Step 2: Draft the context file content

Use assets/prompt-template.md as the structural template — fill in every section with project-specific content discovered in Step 1.

The goal: an AI assistant that has only read this file should accurately answer any question a new developer would ask on day 1:

• "What is this project and who uses it?"
• "What framework and version does it use?"
• "How is the content organized?"
• "What are the naming conventions and frontmatter fields?"
• "How does API authentication work?"

Required sections: Project Overview, Tech Stack, Repository Structure, Content Architecture, Coding Conventions, Common Tasks, Important Files, Do Not. Add API / Integration Patterns if the project documents an API.

Concrete example — a good Project Overview section:

## Project Overview
ShipEdge provides warehouse and order management tools for 3PLs and enterprise e-commerce.
This documentation covers the WMS, OMS, and Returns modules.

**Type:** API Reference + Product Docs
**Audience:** Developers and warehouse operators

Step 3: Create the files

Primary: Create CLAUDE.md at project root (works for Claude Code and is human-readable):

{project_root}/CLAUDE.md

Secondary: Create .cursorrules at project root (works for Cursor):

{project_root}/.cursorrules

Both files should have the same content. The .cursorrules file may use slightly more terse language since Cursor treats it as a system prompt.

Optional: If the project uses GitHub, also create:

{project_root}/.github/copilot-instructions.md

Step 4: Link from docs

For maximum discoverability, add a section to the docs (or README) pointing to these files:

## AI Coding Tools

Context files for AI-assisted development are available:
- [CLAUDE.md](./CLAUDE.md) — for Claude Code
- [.cursorrules](./.cursorrules) — for Cursor

If the project has a "Getting Started for Developers" page, add a note there.

Step 5: Report

Tell the user:

1. Which files were created and where
2. Key sections included
3. How to verify: "open .cursorrules in Cursor and ask 'What is this project?' — it should give an accurate answer"
4. Suggest updating the files whenever major API changes or restructuring occurs

---

Quality Check

A good AI context file passes this test: An AI assistant that has only read this file should be able to accurately answer questions a new developer would ask on day 1.

Test questions:

• "How do I add a new documentation page?"
• "What frontmatter fields does a doc page need?"
• "How does the API authentication work?"
• "What's the URL pattern for [section]?"