openfront-airline-custom-storefront

Start by inspecting the features/storefront directory, which contains the screens, data loaders, sessions, and UI composition. Then decide whether to adapt the built-in client in place or build a separately deployed client against the pinned GraphQL schema.

Adapt the built-in client when you can reuse existing sessions, transport, design tokens, and named server operations. Build a separate client when you need independent deployment, but keep privileged credentials server-side behind a BFF and never expose an arbitrary GraphQL proxy.

Enumerate actual app routes, trace every data loader and server action, list GraphQL operations, and read schema.graphql plus model and field access rules. Record actor, tenant scope, idempotency, and error states for each required operation before proposing code.

No. The Skill explicitly forbids borrowing Ecommerce roots or renaming products and orders into the airline vertical without validating they exist and are authorized. Airline workflows use shopping, booking, check-in, baggage, and disruption concepts with their own authority boundaries.

The Skill requires one bounded server projection or command validating actor, scope, ownership, state transition, and idempotency for consequential actions. If it does not exist, you request the backend contract and show the feature as unavailable rather than faking it in the client.