sf-testing

Run `sf apex run test --class-names MyTest --code-coverage --result-format json --target-org your-alias` to execute a test class with coverage. Use `--test-level RunLocalTests` to run all tests except managed packages, and `--output-dir` to save results for parsing.

The agentic test-fix loop parses failure messages, classifies the error type such as NullPointerException or DmlException, identifies the root cause, delegates the code fix to the sf-apex skill, and re-runs the failing test. It iterates up to three attempts until tests pass.

Salesforce requires a minimum of 75% org-wide Apex coverage for production deployment. Best practice targets 90% or higher overall, with 100% coverage on business-critical code paths.

Salesforce blocks real HTTP requests in tests, so you must implement the HttpCalloutMock interface and register it with Test.setMock before invoking the callout. The mock returns a predefined HttpResponse with your chosen status code and body.

Triggers process records in batches of 200, so testing with 251 records crosses the batch boundary and exposes bulkification bugs like DML or SOQL inside loops. It also verifies the code stays within governor limits across multiple batches.

Use sf-apex when writing or refactoring production Apex code rather than running tests. Use sf-ai-agentforce-testing for Agentforce agent testing and sf-lwc for Jest-based Lightning Web Component tests, as those fall outside Apex test execution scope.