Building Rask apps with AI assistants

Rask ships first-class guidance for AI coding tools, so an assistant can scaffold and extend a Rask app correctly without you re-explaining the conventions.

What's included

• AGENTS.md (repo root) — the AI guidance that ships with the Rask repository. It's the cross-tool standard most AI coding assistants read automatically; it captures the rules that make Rask code compile (the chain not new, the children indexer, step-vs-setter props, the full-shell root, routing/lifecycle, scoped CSS/TypeScript, callbacks, forms, auth). Generated projects no longer ship their own AGENTS.md — point your assistant at this repo-root guidance (and llms.txt).
• llms.txt (repo root) — the emerging standard index that points AI tools at the docs.
• The docs/ set — a task guide for each subsystem (getting-started, elements & the DSL, routing, lifecycle, composition, forms, js-interop, browser APIs, authentication, data access, HTTP & files, PWA, CQRS, diagnostics, testing, … — the full curated list is in the on-site guides index) plus the Tailwind, compiled at build time (docs/tailwind.md: utilities scanned from your own C# source, zero-JS interactivity, typed utility classes). Each guide embeds its examples as live demos, so the source a user reads on GitHub and the running showcase stay in lockstep.

How to use it

1. Scaffold: rask new MyApp.
2. Point your assistant at Rask's repo-root AGENTS.md (and llms.txt if it fetches docs).
3. Ask for features in plain language — the assistant follows the conventions and links to docs/diagnostics.md when it hits a RASKxxx compile diagnostic.
4. Ship it: rask deploy --host user@box --domain app.example.com builds and runs it on a single host over SSH with automatic HTTPS.

Keeping it accurate

The repo-root AGENTS.md and llms.txt are part of the public API surface: when a user-facing behavior changes, they're updated in the same PR (see docs/development-workflow.md). GitHub is the single source of truth — these files are committed, not local-only.