salesforce-apex-test-generator-skill

salesforce-apex-test-generator-skill

T0 static code generation for Apex test classes. This skill authors test coverage that is bulkified, assertion-rich, and factory-patterned — the three qualities most often missing from hand-rolled Apex tests. No org connection required.

When This Skill Owns the Task

Use salesforce-apex-test-generator-skill when the work requires authoring new test classes:

• "Write tests for the AccountService class"
• "Generate a @isTest class for my Queueable job"
• "Create test setup data using TestDataFactory"
• "Add positive, negative, and bulk test methods for OrderTriggerHandler"
• "Scaffold an @TestSetup method for the Contact selector tests"
• "Write async test patterns for my batch class"

Delegate elsewhere when:

Situation	Skill to use
User wants to run the tests against a live org	salesforce-apex-test-runner-skill
User needs the production Apex class authored first	salesforce-apex-generator-skill
User is analyzing a test failure from logs	salesforce-apex-log-analyzer-skill
User needs to deploy validated code	salesforce-deployment-validator-skill

---

Required Context to Gather First

Before generating, confirm:

1. Class under test — the production Apex class name and its purpose
2. Class type — service, selector, domain, batch, queueable, schedulable, invocable, REST resource, trigger
3. SObjects involved — what objects the production class creates, reads, or updates
4. TestDataFactory availability — does the project already have a TestDataFactory class? If yes, use it. If no, scaffold a minimal factory or factory pattern.
5. Key behaviors to test — what are the main positive paths, exception paths, and governor-limit edge cases?
6. API version — default 62.0 minimum

If the user provides the production class source, read it to infer what test scenarios are needed. If not, ask for the class name and its purpose.

---

Recommended Workflow

Step 1 — Read the production class (if available)

Use Read or Grep to locate the class under test. Identify:

• Public methods that need coverage
• Exception paths (try/catch blocks, explicit throws)
• DML operations that need TestDataFactory data
• Async paths (Queueable, Batch, Schedulable enqueue/execute calls)
• callout paths (HTTP callout mocks needed?)

Step 2 — Design test structure

Map each method to test scenarios using the three-path model:

Path	What to test
Positive	Happy path with valid data; asserts on expected outcome
Negative	Invalid input, missing required data, exception thrown; asserts on expected error
Bulk	200+ record insert/update/delete; asserts governor-limit safety

Step 3 — Plan TestDataFactory usage

Consult references/test-data-factory.md. If a project factory exists, extend it. If not, scaffold factory methods as static helpers in the test class or a companion TestDataFactory class.

Step 4 — Author the test class

Follow all hard-stop constraints (see Rules section). Generate:

• @isTest class declaration with private access
• @TestSetup for shared data (when multiple methods need the same records)
• Separate @isTest methods for each positive, negative, and bulk scenario
• Test.startTest / Test.stopTest wrapping all async enqueue or execute calls
• Assert class usage throughout (never bare System.assert)
• No SeeAllData=true unless explicitly required by an external tool pattern

Step 5 — Generate metadata XML

Produce {ClassName}Test.cls-meta.xml with correct apiVersion and status: Active.

Step 6 — Score against the quality rubric

Run the 100-point rubric (see below). If score is below 80, revise before presenting.

Step 7 — Recommend test runner pairing

Always end with an explicit recommendation to execute the tests using salesforce-apex-test-runner-skill, noting the test class name and expected coverage.

---

Rules

Hard-Stop Constraints (Must Enforce)

Constraint	Rationale
Never use @isTest(SeeAllData=true) unless testing external tools that require live data	Test isolation; prevents reliance on org-specific state
Every test method must have at least one Assert statement	Tests without assertions give false confidence
Bulk test methods must use 200+ records	Triggers process in batches of 200; 201 records crosses the boundary
Use Assert class methods (Assert.areEqual, Assert.isTrue, etc.) not bare System.assert	Produces meaningful failure messages; follows Apex best practices
Use Test.startTest / Test.stopTest around all async enqueue or execute calls	Flushes async queue; required for governor-limit isolation in tests
All DML in test setup must use insert not Database.insert unless testing saveResults	Clarity; reserve Database.insert for tests specifically checking partial success
Do not hardcode record Ids in test assertions	Ids vary by org; use the returned record.Id
Separate positive, negative, and bulk methods — do not combine in one method	Maintainability and clarity of failure attribution

Naming Conventions

Element	Pattern	Example
Test class	{ProductionClass}Test	AccountServiceTest
Positive test	test{Method}Success or test{Scenario}	testGetAccountByOwnerSuccess
Negative test	test{Method}ThrowsWhenInvalid or test{Scenario}Error	testGetAccountByOwnerThrowsWhenNullId
Bulk test	test{Method}Bulk or test{Scenario}Bulk	testGetAccountByOwnerBulk
@TestSetup	setup (conventional)	static void setup

---

Quality Scoring Rubric (100-point)

Score the test class before presenting. Threshold: 80+ pass, 60–79 caveat, below 60 revise.

Dimension	Points	What earns full marks
Bulkification	25	At least one bulk test method with 200+ records; asserts on all bulk records not just first
Positive/negative/bulk coverage	25	All three paths present and distinct; each method has a clear purpose
TestDataFactory used	15	Factory methods used or scaffolded; no inline SObject construction scattered across methods
Assert class used	15	Assert.areEqual, Assert.isTrue, Assert.isNotNull used throughout; no bare System.assert(condition)
Governor-limit aware	10	Test.startTest / Test.stopTest wraps async paths; no unnecessary SOQL in each test method
No SeeAllData	10	@isTest(SeeAllData=true) is absent unless explicitly justified with a comment

Scoring penalties:

• Missing bulk test: -20
• Bare System.assert without message: -10 per occurrence (max -20)
• SeeAllData=true without justification: -15
• No Test.startTest on async path: -10
• Assertions only on first record in bulk test: -10
• Missing @TestSetup when multiple methods share data setup: -5

---

T0 Contract

This skill operates exclusively at T0 — static generation only.

• No org connection: No sf CLI calls, no MCP tool calls, no live execution.
• No OAuth required: Zero-scope, zero-credential, zero-network.
• Output is draft code: All generated test classes are starting points for review. Coverage percentages are estimated, not measured. Use salesforce-apex-test-runner-skill for actual execution and coverage measurement.

---

Refusal Triggers

Stop and do not generate if:

• The user requests SeeAllData=true for a standard use case — explain the isolation risk and generate test-isolated factory data instead. Only emit SeeAllData=true if the user provides a documented reason (e.g., testing a legacy package that requires live data).
• The user requests test methods that assert on hardcoded record Ids — explain the org-portability risk and generate assertions using returned record Ids instead.

---

Output Format

verdict: "pass | caveat | reject"
quality_score: <0-100>
quality_notes: "<scoring rationale>"

generated_files:
  - path: "{ClassName}Test.cls"
    content: |
      <apex test code>
  - path: "{ClassName}Test.cls-meta.xml"
    content: |
      <meta xml>

test_scenarios_covered:
  positive: ["<method name: scenario description>"]
  negative: ["<method name: scenario description>"]
  bulk: ["<method name: scenario description>"]

factory_approach: "existing-factory | scaffolded-factory | inline-helpers"
factory_notes: "<what factory methods were used or created>"

estimated_coverage_areas:
  - "<class/method: coverage area>"

async_patterns_used: ["Test.startTest/stopTest", "callout mock", "schedulable mock"]

runner_recommendation:
  companion_skill: "salesforce-apex-test-runner-skill"
  test_class: "{ClassName}Test"
  expected_coverage_target: ">=75%"

assumptions:
  - "<list of assumptions made>"

missing_context:
  - "<what the user should provide to improve coverage>"

---

Handoff Rules

Output	Hand off to
Generated test class ready for execution	salesforce-apex-test-runner-skill (T1)
Production class still needed	salesforce-apex-generator-skill
Test failures need log analysis	salesforce-apex-log-analyzer-skill
Ready for sandbox deploy	salesforce-deployment-validator-skill (T2)

---

Stop Conditions

Stop and do not continue if:

• The production class under test cannot be located and the user has not provided its source or purpose — ask for the class content or a description of its methods.
• The requested test patterns would require live org data not expressible via factory records — explain the limitation and recommend an approach.

---

Security Notes

• T0 static generation only: No org connection, no OAuth, no secrets.
• No SeeAllData by default: Test isolation is enforced. Live org data access in tests is a security and reliability anti-pattern.
• No credential generation: Test classes never contain hardcoded credentials, org Ids, or session tokens.
• Factory data only: All test data is generated via factory methods in the test class or TestDataFactory. No reliance on existing org records.

---

Reference File Index

File	When to read
references/test-data-factory.md	TestDataFactory pattern, builder pattern, field overrides, lazy init, duplicate rule handling
references/assertion-patterns.md	Assert class methods, error messages, multi-assert vs single, bulk assertion patterns
references/async-testing.md	Test.startTest/stopTest, mocking Schedulable/Batchable, HTTP callout mocks