FAQ

Do I need to know YAML? No. The /fdl-create and /fdl-extract commands handle everything through plain-language questions. You only see YAML if you choose to edit blueprints directly.

Can I extract from an existing codebase? Yes. /fdl-extract-code reads a local folder or clones a git repo, then analyzes models, routes, middleware, validators, services, error handling, events, and tests. Works with any tech stack. For large repos, use /fdl-extract-code-feature to select only the features you want.

Does this only work with Claude? No. The blueprints are standard YAML files — any AI tool can read them. You can paste a blueprint into ChatGPT, Copilot, Gemini, or any other AI. The slash commands are Claude Code skills that make the experience smoother, but the blueprints themselves are AI-agnostic. The JSON API makes blueprints even easier for AI tools to fetch.

What languages and frameworks are supported? All of them. Blueprints describe what the feature does, not how to build it. Next.js, Express, Laravel, Angular, React, Vue, C#/.NET, Rust, Python/Django, Go, Ruby on Rails, Flutter, Swift, and anything else.

Can I use this for business processes, not just UI features? Yes. Blueprints support actors, state machines, SLAs, and approval chains. See the expense-approval blueprint for a full example.

Can I extract rules from existing documents? Yes. /fdl-extract reads PDFs, Word docs, text files, and even images of flowcharts.

Can I extract from a website or API docs? Yes. /fdl-extract-web crawls documentation websites (even JS-rendered ones) using Chrome. It discovers OpenAPI specs and Postman collections automatically.

What if I don’t know what feature I want? Use /fdl-brainstorm. Describe your problem in plain language and Claude walks you through structured questions — what problem it solves, what success looks like, what can go wrong — until the feature shape is clear. It then hands off to /fdl-create with a complete spec.

Can I use blueprints with ChatGPT or Gemini directly? Yes. Run /fdl-build-yaml with your app description — it exports a self-contained YAML build pack that any AI tool can use. Choose full, chunked (for small context windows), or compact (~60% smaller) format. Upload the file to ChatGPT, Gemini, or any AI and it generates code from the blueprints.

How do I set up FDL in Cursor, Windsurf, or Copilot? Run /fdl-install cursor (or windsurf, copilot, gemini, continue, cline, kiro, amazonq, codex). It writes the right config file for your tool so it knows how to find and use blueprints. Supports 9 AI coding tools.

Do I need to manually validate and commit after changes? No. /fdl-auto-evolve handles it automatically — validates all blueprints, regenerates docs and the JSON API, and creates a single atomic commit. It runs automatically after /fdl-create and /fdl-extract-code, or you can trigger it manually.

How is this different from just asking AI to “build login”? Without FDL, the AI guesses. With FDL, there’s a complete specification: 5 failed attempts = lockout, 15-minute duration, constant-time password comparison, generic error messages. Nothing is left to chance.