platform-apex-test-run: Salesforce Test Execution & Coverage Analysis
Use this skill when the user needs Apex test execution and failure analysis: running tests, checking coverage, interpreting failures, improving coverage, and managing a disciplined test-fix loop for Salesforce code.
When This Skill Owns the Task
Use platform-apex-test-run when the work involves:
sf apex run test workflows- Apex unit-test failures
- code coverage analysis
- identifying uncovered lines and missing test scenarios
- structured test-fix loops for Apex code
Delegate elsewhere when the user is:
- writing or refactoring production Apex →
platform-apex-generate skill
- testing Agentforce agents →
agentforce-test skill
- testing LWC with Jest → experience-lwc-generate
Required Context to Gather First
Ask for or infer:
- target org alias
- desired test scope: single class, specific methods, suite, or local tests
- coverage threshold expectation
- whether the user wants diagnosis only or a test-fix loop
- whether related test data factories already exist
Recommended Workflow
1. Discover test scope
Identify:
- existing test classes
- target production classes
- test data factories / setup helpers
2. Run the smallest useful test set first
Start narrow when debugging a failure; widen only after the fix is stable.
3. Analyze results
Focus on:
- failing methods
- exception types and stack traces
- uncovered lines / weak coverage areas
- whether failures indicate bad test data, brittle assertions, or broken production logic
4. Run a disciplined fix loop
When the issue is code or test quality:
- delegate code fixes to
platform-apex-generate skill when needed
- add or improve tests
- rerun focused tests before broader regression
5. Improve coverage intentionally
Cover:
- positive path
- negative / exception path
- bulk path (251+ records where appropriate)
- callout or async path when relevant
High-Signal Rules
| Rule | Rationale |
|---|
Default to SeeAllData=false | Ensures test isolation; prevents reliance on org-specific data |
| Every test must assert meaningful outcomes | Tests with no assertions prove nothing and give false confidence |
| Test bulk behavior with 251+ records | Triggers process in batches of 200; 251 records crosses the boundary |
Use factories / @TestSetup when they improve clarity | Consistent data creation in one place; rolled back between test methods |
Pair Test.startTest() with Test.stopTest() for async | Ensures async operations (queueable, future) complete before assertions |
| Do not hide flaky org dependencies inside tests | Prevents intermittent failures tied to org state |
Gotchas
| Issue | Resolution |
|---|
| Test passes locally but fails in CI org | Check for SeeAllData=true or undeclared dependencies on org-specific records |
| Coverage drops unexpectedly after refactor | Run focused class-level tests first, then widen to RunLocalTests to confirm |
| "Uncommitted work pending" error in callout test | DML and HTTP callouts cannot be mixed in the same test context without Test.startTest() wrapping |
| Mock not taking effect in test | Ensure Test.setMock() is called before the code that makes the callout |
@TestSetup data missing in test method | @TestSetup data is committed per test method — re-query it; do not store in static variables |
| API version 67.0 and higher without necessary access level checks | Check failing SOQL/DML stack traces for CRUD/FLS access errors, using System.runAs with an assigned permission set when user-mode behavior is intended, or documenting a justified SYSTEM_MODE path when system access is required |
Output Format
When finishing, report in this order:
- What tests were run
- Pass/fail summary
- Coverage result
- Root-cause findings
- Fix or next-run recommendation
Suggested shape:
Test run: <scope>
Org: <alias>
Result: <passed / partial / failed>
Coverage: <percent / key classes>
Issues: <highest-signal failures>
Next step: <fix class, add test, rerun scope, or widen regression>
Cross-Skill Integration
| Need | Delegate to | Reason |
|---|
| Fix production code or author test classes | platform-apex-generate skill | Code generation and repair |
| Create bulk / edge-case test data | platform-data-manage | Realistic test datasets |
| Deploy updated tests to org | platform-metadata-deploy | Deployment workflows |
| Inspect detailed runtime logs | platform-apex-logs-debug | Deeper failure analysis |
Reference File Index
| File | When to read |
|---|
references/cli-commands.md | All sf apex run test command flags, output formats, async execution, and coverage commands |
references/test-patterns.md | Test class templates — basic, bulk (251+), mock callout, and data factory patterns |
references/testing-best-practices.md | Core testing principles — AAA pattern, naming conventions, bulk, negative, and mock strategies |
references/test-fix-loop.md | Agentic test-fix loop implementation and failure analysis decision tree |
references/mocking-patterns.md | HttpCalloutMock, DML mocking, StubProvider, and selector mocking patterns |
references/performance-optimization.md | Techniques to reduce test execution time — DML mocking, SOQL mocking, loop optimizations |
assets/basic-test.cls | Template: standard test class with @TestSetup, positive / negative / bulk / edge-case methods |
assets/bulk-test.cls | Template: bulk test with 251+ records that crosses the 200-record trigger batch boundary |
assets/mock-callout-test.cls | Template: HTTP callout mock using HttpCalloutMock |
assets/test-data-factory.cls | Template: reusable TestDataFactory with create and insert helpers |
assets/dml-mock.cls | Template: IDML interface + DMLMock implementation for database-free unit tests |
assets/stub-provider-example.cls | Template: StubProvider-based dependency injection stub |
scripts/parse-test-results.py | Post-tool hook — parses sf apex run test JSON output and formats failures for the auto-fix loop |
Score Guide
| Score | Meaning |
|---|
| 108+ | strong production-grade test confidence |
| 96–107 | good test suite with minor gaps |
| 84–95 | acceptable but strengthen coverage / assertions |
| < 84 | below standard; revise before relying on it |
