LogoDocs
Docs/Skill files

Skill files

Skill files are the core of the build package. Each one is a focused instruction document that teaches your AI exactly how to implement one domain of your app — from data models to edge cases to the specific libraries to use.

What is a skill file?

A skill file is a markdown document placed in /skills/. Your AI IDE reads all of them before writing any code. Think of each file as a senior engineer's handoff note for one part of the system.

Unlike a generic README or documentation, skill files are generated specifically for your app — they reference your data models, your user persona from the validation, and the decisions made in project.md.

ℹ️
Skill files don't contain code. They contain decisions, patterns, and constraints that guide the AI as it writes code. This keeps them reusable across sessions and easy to edit.

What's in each file

Every skill file follows the same structure:

  • Overview — what this domain does in this app
  • Stack decisions — which libraries/services to use and why
  • Data models — tables, types, and relationships relevant to this domain
  • Implementation patterns — how to structure the code for this domain
  • Edge cases — gotchas, error states, and things to handle explicitly
  • What NOT to do — anti-patterns to avoid

The skill files generated for your app

skill-core-app.md

Always included. Covers the main app shell — Next.js app router structure, shared layouts, navigation, loading states, error boundaries, and global conventions like naming, file organization, and component patterns.

skill-auth.md

Supabase auth setup, session management, protected routes, middleware, the sign-in/sign-up flow, email confirmation, password reset, and OAuth providers. Includes the user profile table pattern.

skill-billing.md

Stripe integration, subscription plans, usage gating, webhook handling, the customer portal, and how to gate features behind plan level. Includes the Supabase subscription sync pattern.

Domain-specific skills (2–4 files)

The platform generates additional skill files based on what your specific app needs. Examples:

  • skill-notifications.md — email/push delivery, queuing, preferences
  • skill-search.md — full-text search, filters, pagination patterns
  • skill-ai-integration.md — prompt patterns, streaming responses, cost controls
  • skill-file-uploads.md — Supabase Storage, size limits, type validation
  • skill-analytics.md — event tracking, dashboards, retention metrics
  • skill-api.md — public API design, rate limiting, key management

Adding your own skill files

Drop any skill-*.md file into the /skills/ folder. As long as CLAUDE.md instructs the AI to read everything in /skills/, it will pick it up automatically in the next session.

Use this to add context the generator didn't know about — a third-party API you're integrating, a specific design system, company coding conventions, or a domain the AI keeps getting wrong.

markdown
# skill-my-integration.md

## Overview
This app integrates with the Acme API for X.

## Auth
Use API key stored in ACME_API_KEY env var.
Never expose this client-side.

## Key endpoints
- POST /jobs — create a processing job
- GET /jobs/:id — poll for completion

## Error handling
Acme returns 429 on rate limit. Implement
exponential backoff with max 3 retries.

## What NOT to do
Do not use the webhook endpoint — it requires
a paid plan we don't have yet.
💡
Keep skill files focusedOne domain per file. If a skill file is growing beyond 300 lines, split it. The AI reads better when each file has a single clear purpose.