Fix Tests

NeoLabHQ/context-engineering-kit/skills/fix-tests

作者 NeoLabHQ23e2428e809d77717f8acc9659c374a3a1fcb93e无许可证收录于 2026年10月9日更新于 2026年10月9日

Systematically fix all failing tests after business logic changes or refactoring

AI 生成的概览

编排智能体在业务逻辑变更后修复失败的测试,保持测试意图与覆盖率。

功能
该技能定义了一套在业务逻辑变更、重构或依赖更新后修复失败测试的工作流程。它会发现测试基础设施,运行完整测试套件以找出失败文件,验证单个测试文件能否独立运行,然后为每个失败的测试文件启动一个 developer 智能体来修复断言、测试环境或 mock。它会重新运行完整套件并反复迭代,直到所有测试通过,除发现真实缺陷外不改动业务逻辑。
适用场景
适用于测试因不再匹配当前行为或实现而失败的情况,例如重构或依赖更新之后。它旨在修复测试而非修改业务逻辑,并可通过用户参数限定到特定测试或模块。
运行要求
仅以指令形式运行,不附带脚本。它需要一个可运行测试套件的项目以及 README.md、package.json 等配置文件,并依赖智能体平台启动子智能体的能力。若可用,它还会可选地读取用于智能体管理的 sadd 技能和用于测试编写实践的 TDD 技能。

Fix Tests

User Arguments

User can provide to focus on specific tests or modules:

$ARGUMENTS

If nothing is provided, focus on all tests.

Context

After business logic changes, refactoring, or dependency updates, tests may fail because they no longer match the current behavior or implementation. This command orchestrates automated fixing of all failing tests using specialized agents.

Goal

Fix all failing tests to match current business logic and implementation.

Important Constraints

  • Focus on fixing tests - avoid changing business logic unless absolutely necessary
  • Preserve test intent - ensure tests still validate the expected behavior
  • "Analyse complexity of changes" -
    • if there 2 or more changed files, or one file with complex logic, then Do not write tests yourself - only orchestrate agents!
    • if there is only one changed file, and it's a simple change, then you can write tests yourself.

Workflow Steps

Preparation

  1. Read sadd skill if available

    • If available, read the sadd skill to understand best practices for managing agents
  2. Discover test infrastructure

    • Read @README.md and package.json (or equivalent project config)
    • Identify commands to run tests and coverage reports
    • Understand project structure and testing conventions
  3. Run all tests

    • Execute full test suite to establish baseline
  4. Identify all failing test files

    • Parse test output to get list of failing test files
    • Group by file for parallel agent execution

Analysis

  1. Verify single test execution
    • Choose any test file
    • Launch haiku agent with instructions to find proper command to run this only test file
      • Ask him to iterate until you can reliably run individual tests
    • After he complete try running a specific test file if it exists
    • This ensures agents can run tests in isolation

Test Fixing

  1. Launch developer agents (parallel)

    • Launch one agent per failing test file
    • Provide each agent with clear instructions:
      • Context: Why this test needs fixing (business logic changed)
      • Target: Which specific file to fix
      • Guidance: Read TDD skill (if available) for best practices how to write tests.
      • Resources: Read README and relevant documentation
      • Command: How to run this specific test file
      • Goal: Iterate until test passes
      • Constraint: Fix test, not business logic (unless clearly broken)
  2. Verify all fixes

    • After all agents complete, run full test suite again
    • Verify all tests pass
  3. Iterate if needed

    • If any tests still fail: Return to step 5
    • Launch new agents only for remaining failures
    • Continue until 100% pass rate

Success Criteria

  • All tests pass ✅
  • Test coverage maintained
  • Test intent preserved
  • Business logic unchanged (unless bugs found)

Agent Instructions Template

When launching agents, use this template:

The business logic has changed and test file {FILE_PATH} is now failing.
Your task:1. Read the test file and understand what it's testing2. Read TDD skill (if available) for best practices on writing tests.3. Read @README.md for project context4. Run the test: {TEST_COMMAND}5. Analyze the failure - is it:   - Test expectations outdated? → Fix test assertions   - Test setup broken? → Fix test setup/mocks   - Business logic bug? → Fix logic (rare case)6. Fix the test and verify it passes7. Iterate until test passes

来源与署名

来源:NeoLabHQ/context-engineering-kit位于skills/fix-tests提交23e2428

许可证: 无许可证

内容归原作者所有。SourceWeft 从公开仓库中收录这些内容。

举报或申请下架