qt-unittest-make

Qt Unit Test Generator

Overview

Core principle: Use LSP to precisely analyze the class structure, generate test cases with 100% function coverage, and enforce verification that the build succeeds.

Architecture Pattern: Skill routing + sub-agent full-stack execution. Sub-agents are responsible for all concrete work; the Skill is only responsible for invocation and result feedback.

Key Features: Intelligent CMake merging, incremental update support, strict build verification, detailed error handling.

Iron Laws

1. Use Google Test only: The test framework is fixed to GTest; Qt Test is not supported
2. 100% function coverage: Each public/protected function must have at least one test case
3. Intelligent CMake merging: Optimize the merge according to the actual project situation to ensure generality
4. Support incremental updates: Compare with existing tests and fill in uncovered functions
5. Must verify the build: After generation, it must compile successfully before completion can be reported
6. Compilation failures must be fixed: Retry each error at most 3 times, with a maximum of 10 loops

When to Use

Trigger Conditions:

• The user requests unit tests for a module:为 src/lib/ui 模块创建单元测试
• The user requests tests for a class:为 MyClass 生成测试
• The user requests test completion:为 MyClass 补全测试

When not to use this skill:

• The user requests generating a test framework (use qt-unittest-build）
• The user requests fixing test failures (use systematic-debugging）
• The user requests analyzing test coverage (use other tools)

Execution Flow

Step 1: Check the test framework

Verify that autotests/CMakeLists.txt and autotests/3rdparty/stub/ exist. If they do not exist, prompt the user to first run qt-unittest-build。

Step 2: Analyze the class structure

Batch module generation: Glob-scan all .h/.hppin the directory and extract class names.

Single class: Directly analyze the specified class.

Use LSP tools:

• lsp_document_symbols - Extract class structure
• lsp_goto_definition - Read function implementations
• lsp_find_references - Find dependencies

Step 3: Invoke sub-agents (MUST DO)

Important: Sub-agents must be invoked; do not skip this!

Sub-agent location: agent/unittest-generator.md

Invocation method:

Use the task tool to invoke sub-agents:

task(
    description="生成单元测试代码",
    prompt="完整的任务说明，包括：
- 目标模块或类
- 测试框架要求（Google Test）
- 函数覆盖率要求（100%）
- 验证构建要求",
    subagent_type="general"
)

Why sub-agents are needed:

• Independent context: When generating tests for a single class or small module, an independent context avoids contamination
• Parallel execution: When batch-generating tests for multiple classes, multiple sub-agents can be forked to execute in parallel, improving efficiency
• Task isolation: A sub-agent failure does not affect the main Agent, making error handling easier

Invocation timing:

• Batch module generation: Invoke one sub-agent for each class (in parallel)
• Single class generation: Invoke one sub-agent
• Incremental update: Invoke a sub-agent to analyze the differences

Step 4: Wait for sub-agents to complete

Monitor the sub-agents' execution results:

• Collect the generated test files
• Check the verification build results
• Record any errors encountered (if any)

Step 5: Provide feedback to the user

Provide feedback to the user based on the sub-agents' results:

• If successful: report the generated test files and coverage
• If failed: report detailed error information and correction suggestions

Red Flags

• ❌ Use the Qt Test framework
• ❌ Coverage less than 100%
• ❌ Hardcoded CMake template
• ❌ Do not verify the build
• ❌ Do not support incremental updates
• ❌ Report completion even if compilation fails
• ❌ Skip sub-Agent invocation (work directly by hand)

Quick Reference

Test file naming: test_myclass.cpp

Test class naming: MyClassTest

Test case naming: {Feature}_{Scenario}_{ExpectedResult}

LSP tools: lsp_document_symbols, lsp_goto_definition, lsp_find_references

Stub mode: &Class::method, VADDR(Class, method), static_cast<...>

Compilation retry logic: retry each error up to 3 times, with a maximum of 10 loops

Sub-Agent invocation

The task tool must be used to invoke the sub-Agent:

task(
    description="生成单元测试代码",
    prompt="完整任务描述，包括目标、要求、验证流程",
    subagent_type="general"
)

Sub-Agent execution content:

• Analyze the project structure (LSP)
• Generate test files (100% coverage)
• Intelligently merge CMake
• Verify the build (retry each error 3 times, with a maximum of 10 loops)

Why a sub-Agent is needed:

• Independent context to prevent contamination
• Parallel execution to improve efficiency
• Task isolation for easier error handling

Common errors

Error	Cause	Fix
Test framework does not exist	qt-unittest-build was not run	Prompt the user to run the framework build skill first
Insufficient coverage	Not all functions were analyzed	Ensure lsp_document_symbols extracts completely
CMake merge failed	Hardcoded template	Use AI intelligent merging and optimize according to the actual project situation
Compilation failed	Stub signature error	Use LSP to obtain precise signatures
Report completion even if compilation fails	Skipped verification or verification was not rigorous	Compilation must succeed before reporting to the user
Multiple errors were not corrected	Global retry 3 times is insufficient	Retry each error 3 times, with a maximum of 10 loops

Rationalization Table

Excuse	Reality
"Working directly by hand is faster"	Manual work cannot guarantee 100% coverage and build verification; a sub-Agent is mandatory
"Skip sub-Agent invocation"	This violates the iron rule of the skill; the sub-Agent must be invoked
"Completion can also be reported if compilation fails"	This violates the mandatory build verification requirement; compilation must succeed before reporting
"Global retry 3 times is enough"	Retry each error 3 times, not 3 times globally
"80% coverage is enough"	The iron rule requires 100% function coverage, with no exceptions
"Qt Test is the same"	The skill uses Google Test exclusively and does not support Qt Test