deploying-ui-bundle

A deployment-sequencing skill for shipping a UI bundle application to a Salesforce org, where the order of operations is critical. It activates when a project contains a `uiBundles/*/src/` directory or `sfdx-project.json` and the task involves deploying, pushing to an org, or post-deploy setup, including when files like `*.uibundle-meta.xml` or `sfdx-project.json` exist and the user mentions deploying or org setup. It lays out the canonical end-to-end flow from authentication through final build.

The sequence begins by checking and, if needed, authenticating the org, since all later steps require it. Next it installs dependencies and builds the UI bundle to produce `dist/`, required before deploying UI bundle entities. Metadata deployment then checks for a manifest (`manifest/package.xml` or `package.xml`) and deploys via the manifest if present, otherwise deploying all project metadata—objects, layouts, permission sets, Apex classes, UI bundles, and more—which must complete before schema fetch because the schema reflects org state. Post-deploy configuration assigns permission sets and groups (required for GraphQL introspection to return the correct schema), ensures correct profiles, and sets up named credentials, connected apps, custom settings, and flow activation; after a successful deploy it proactively discovers permission sets under `force-app/main/default/permissionsets/` and assigns each one or asks the user.

Optional data import runs only when `data/data-plan.json` exists, deleting in reverse plan order (children before parents) and importing via Anonymous Apex with duplicate-rule save enabled, always with user confirmation. Finally it sets the default org, fetches the GraphQL schema via introspection (writing `schema.graphql` at the project root), runs codegen against the local schema, and builds the UI bundle if not already built. Critical rules emphasize deploying metadata before fetching schema, assigning permissions before schema fetch so field-level security is correct, and re-running schema fetch and codegen after every metadata deployment that changes objects, fields, or permissions—never silently skipping permission assignment or data import.