sf-flow

Start from the before-save or after-save template, define the trigger object and entry conditions, then generate the .flow-meta.xml with alphabetically ordered elements and fault connectors. Use before-save for same-record field updates and after-save for related records, emails, or callouts.

Use Transform when the job is data mapping or reshaping collections, since it runs 30-50% faster than loops. Use a Loop only when you need per-record decision logic or branching that Transform cannot express.

Before-save flows update fields on the triggering record without DML and run fastest. After-save flows handle related record updates, emails, and callouts but consume more resources because they run after the record commits.

Failures usually come from DML statements or Get Records inside loops, which hit governor limits past 200 records. Move DML outside loops, use collection variables, and test with 251 or more records to cross the batch boundary.

Yes, but deploying the Flow alone is not enough. You must register it under Setup > Agentforce > Action Definitions, map input and output variables, and ensure variable API names match the Agent Script exactly.

Use sf flow run test --target-org <alias> --json to execute all flow tests, or pass --class-names for a specific flow. Retrieve asynchronous results with sf flow get test --test-run-id <id>.