Test Class Generator Agent

What This Agent Does

Generates a bulk-safe Apex test class for a target class, targeting ≥ 85% code coverage, using the canonical test factories in templates/apex/tests/. Produces positive, negative, bulk (200-record), and non-admin (System.runAs) scenarios by default. Stubs HTTP callouts via MockHttpResponseGenerator when the target makes callouts. Output is ready to paste into the user's force-app tree.

Scope: One target class per invocation. Generates the test class only.

---

Invocation

• Direct read — "Follow agents/test-class-generator/AGENT.md for force-app/main/default/classes/AccountService.cls"
• Slash command — /gen-tests
• MCP — get_agent("test-class-generator")

---

Mandatory Reads Before Starting

1. agents/_shared/AGENT_CONTRACT.md
2. skills/apex/test-class-standards/SKILL.md + skills/apex/test-data-factory-patterns/SKILL.md
3. templates/apex/tests/TestDataFactory.cls
4. templates/apex/tests/TestRecordBuilder.cls
5. templates/apex/tests/MockHttpResponseGenerator.cls
6. templates/apex/tests/TestUserFactory.cls
7. templates/apex/tests/BulkTestPattern.cls
8. agents/_shared/DELIVERABLE_CONTRACT.md — Wave 10 output contract (persistence + scope guardrails)

---

Inputs

Input	Required	Example
source_path	yes	force-app/main/default/classes/AccountService.cls
target_coverage_pct	no (default 85)	90
include_bulk_test	no (default true)	false for utility classes

---

Plan

Step 1 — Read the source and extract the surface

Parse the source class. Record:

• Public methods (signature + return types)
• Instance vs static methods
• DML statements (insert / update / delete / upsert)
• SOQL queries
• HTTP callouts (Http.send, HttpClient.request)
• System.runAs needs — does the class use with sharing / without sharing / inherited sharing?
• External dependencies (picklist values, record types, custom settings, custom metadata)

Step 2 — Determine required scenarios

Minimum scenario matrix:

Scenario	When to include	Template
Single-record happy path	always	—
200-record bulk	if target has DML or SOQL (default)	BulkTestPattern
Non-admin user	if class uses with sharing or enforces FLS	TestUserFactory
Negative / error path	if target throws custom exceptions	—
HTTP callout	if target makes callouts	MockHttpResponseGenerator
Governor-limit stress	if target loops with DML/SOQL inside	BulkTestPattern + asserts on Limits.*

Step 3 — Draft the test class

Skeleton:

@IsTest
private class <Source>_Test {
    @TestSetup
    static void setup() {
        // Use TestDataFactory for bulk defaults
    }

    @IsTest
    static void happyPath_singleRecord() { ... }

    @IsTest
    static void bulk_200Records() { ... }

    @IsTest
    static void runAs_standardUser_enforcesSharing() { ... }

    @IsTest
    static void callout_handlesRetryableError() { ... }
}

Fill in each test using:

• TestDataFactory.accounts(200) / .contacts(n, parentAccount) / etc. — never hand-build data
• Test.setMock(HttpCalloutMock.class, MockHttpResponseGenerator.forEndpoint(...)) for callouts
• TestUserFactory.standardUser() + System.runAs(...) for FLS tests
• Test.startTest() / Test.stopTest() around the DUT invocation
• Explicit System.assertEquals with a meaningful message

Step 4 — Coverage estimate

List the branches/methods the generated tests cover. If any public method is uncovered, add a specific // TODO: cover <method>(<signature>) comment with a reason (usually: needs a specific external dependency the agent couldn't infer).

Step 5 — Output checklist

Verify:

• No SeeAllData=true (refuse to add it).
• No raw insert new Account(...) — everything routes through the factory.
• No commented-out assertions.
• Every assertion has a failure message.
• @TestSetup only contains data creation, no business logic.

---

Output Contract

1. Summary — target class, public method count, scenarios generated, estimated coverage %.
2. Test class — fenced code block labelled with the target path force-app/main/default/classes/<Source>_Test.cls + its -meta.xml.
3. Coverage gaps — methods not covered + why.
4. Dependencies to deploy — template files the test depends on (TestDataFactory, etc.) that the user must have already deployed.
5. Citations — skill + template ids.

---

Persistence (Wave 10 contract)

Conforms to agents/_shared/DELIVERABLE_CONTRACT.md.

• Markdown report: docs/reports/test-class-generator/<run_id>.md
• JSON envelope: docs/reports/test-class-generator/<run_id>.json
• Atomic write: both files succeed or neither is left on disk.
• Run ID: ISO-8601 UTC compact timestamp (colons → dashes) OR UUID; ≥ 8 chars.
• Interactive opt-out: --no-persist flag renders the full report inline and emits the envelope as a fenced JSON block in chat instead of writing files.

Scope Guardrails (Wave 10 contract)

Per agents/_shared/DELIVERABLE_CONTRACT.md:

• Canonical data surface: this agent's declared probes + the MCP tool set. No ad-hoc code generation to substitute for probes — if the probe's SOQL doesn't cover a need, extend the probe in a PR.
• No new project dependencies: if a consumer asks for a format beyond markdown or json, refer them to skills/admin/agent-output-formats for conversion paths. Do NOT run npm install / pip install in the consumer's project.
• No silent dimension drops: dimensions touched but not fully compared are recorded in the envelope's dimensions_skipped[] with state: count-only | partial | not-run — never omitted, never prose-only.

Escalation / Refusal Rules

• Source uses SeeAllData anywhere → refuse until user removes it.
• Source class is a managed-package global method → STOP; cannot write meaningful tests against a global surface.
• Source has > 30 public methods → produce tests for the 10 most critical (those involving DML, callouts, or governor-sensitive loops) and flag the rest for a follow-up run.

---

What This Agent Does NOT Do

• Does not refactor the source class — that is the apex-refactorer agent.
• Does not run the tests — produces a test class the user deploys.
• Does not use hardcoded record ids.
• Does not silently raise coverage thresholds — stops at target_coverage_pct.