MCP HubMCP Hub
返回技能列表

overnight-development

jeremylongshore
更新于 Today
9 次查看
409
51
409
在 GitHub 上查看
aitestingautomationdesign

关于

This Claude Skill automates overnight software development by using Git hooks to enforce test-driven development (TDD). It autonomously builds features, refactors code, and fixes bugs while ensuring all tests pass before allowing commits. Use it when you need autonomous coding that guarantees production-ready, fully tested code.

技能文档

Overnight Development

Overview

This skill automates software development overnight by leveraging Git hooks to enforce test-driven development (TDD). It ensures that all code changes are fully tested and meet specified quality standards before being committed. This approach allows Claude to work autonomously, building new features, refactoring existing code, or fixing bugs while adhering to a rigorous TDD process.

Core Capabilities

  • Enforces test-driven development (TDD) using Git hooks.
  • Automates debugging and code fixing until all tests pass.
  • Tracks progress and logs activities during overnight sessions.
  • Supports flexible configuration for various testing frameworks and languages.
  • Provides guidance and support through the overnight-dev-coach agent.

Workflow

Phase 1: Project Setup and Configuration

To prepare the project for overnight development:

  1. Verify Prerequisites: Ensure the project is a Git repository, has a configured test framework, and includes at least one passing test.

    git init
    npm install --save-dev jest # Example for Node.js
    
  2. Install the Plugin: Add the Claude Code Plugin marketplace and install the overnight-dev plugin.

    /plugin marketplace add jeremylongshore/claude-code-plugins
    /plugin install overnight-dev@claude-code-plugins-plus
    
  3. Run Setup Command: Execute the /overnight-setup command to create necessary Git hooks and configuration files.

    /overnight-setup
    

Phase 2: Task Definition and Planning

To define the task for the overnight session:

  1. Define a Clear Goal: Specify a clear and testable goal for the overnight session, such as "Build user authentication with JWT (90% coverage)."

    Task: Build user authentication with JWT (90% coverage)
    Success: All tests pass, 90%+ coverage, fully documented
    
  2. Start Coding: Begin implementing the feature by writing tests first, following the TDD approach.

    // Example test case (Node.js with Jest)
    it('should authenticate a user with valid credentials', async () => {
      // Test implementation
    });
    
  3. Attempt to Commit: Try to commit the changes, which will trigger the Git hooks and run the tests.

    git commit -m "feat: implement user authentication"
    

Phase 3: Autonomous Development and Debugging

To allow Claude to work autonomously:

  1. Git Hooks Enforcement: The Git hooks will block the commit if any tests fail, providing Claude with the error messages.

    Overnight Dev: Running pre-commit checks...
    Running linting...
    Linting passed
    Running tests...
    12 tests failing
    Commit blocked!
    
  2. Automated Debugging: Claude analyzes the error messages, identifies the issues, and attempts to fix the code.

    Claude: Fixing test failures in user authentication module.
    
  3. Retry Commits: Claude retries the commit after making the necessary fixes, repeating the process until all tests pass.

    git commit -m "fix: address test failures in user authentication"
    

Phase 4: Progress Tracking and Completion

To monitor the progress and finalize the session:

  1. Monitor Progress: Track the progress of the overnight session by viewing the log file.

    cat .overnight-dev-log.txt
    
  2. Review Results: Wake up to fully tested code, complete features, and a clean Git history.

    7 AM: You wake up to:
    - 47 passing tests (0 failing)
    - 94% test coverage
    - Clean conventional commit history
    - Fully documented JWT authentication
    - Production-ready code
    
  3. Session Completion: The session completes when all tests pass, the code meets the specified quality standards, and the changes are committed.

Using Bundled Resources

Scripts

To automate the setup process, use the overnight-setup.sh script:

./scripts/overnight-setup.sh

To track the progress of the overnight session, use the progress-tracker.py script:

./scripts/progress-tracker.py --log .overnight-dev-log.txt

References

For detailed configuration options, load:

For best practices on writing effective tests, load:

Assets

Available templates:

  • assets/commit-template.txt - Template for generating commit messages.
  • assets/readme-template.md - Template for generating README files.

Examples

Example 1: Building JWT Authentication

User request: "Implement JWT authentication with 90% test coverage overnight."

Workflow:

  1. Claude writes failing authentication tests (TDD).
  2. Claude implements JWT signing (tests still failing).
  3. Claude debugs token generation (commit blocked, keeps trying).
  4. Tests pass! Commit succeeds.
  5. Claude adds middleware (writes tests first).
  6. Integration tests (debugging edge cases).
  7. All tests green (Coverage: 94%).
  8. Claude adds docs, refactors, still green.
  9. Session complete.

Example 2: Refactoring Database Layer

User request: "Refactor the database layer to use the repository pattern overnight."

Workflow:

  1. Claude analyzes existing tests to ensure no regression.
  2. Claude implements the repository pattern.
  3. Tests are run; some fail due to changes in data access.
  4. Claude updates tests to align with the new repository pattern.
  5. All tests pass; commit succeeds.
  6. Claude documents the refactored database layer.
  7. Session complete.

Example 3: Fixing a Bug in Payment Processing

User request: "Fix the bug in payment processing that causes incorrect amounts to be charged overnight."

Workflow:

  1. Claude reproduces the bug and writes a failing test case.
  2. Claude analyzes the code and identifies the root cause of the bug.
  3. Claude fixes the bug and runs the tests.
  4. The failing test case now passes; all other tests also pass.
  5. Commit succeeds.
  6. Claude adds a comment to the code explaining the fix.
  7. Session complete.

Best Practices

  • Ensure that the task is well-defined and testable.
  • Follow the TDD approach by writing tests before implementing features.
  • Monitor the progress of the overnight session by viewing the log file.
  • Configure the Git hooks and settings appropriately for the project.
  • Use the overnight-dev-coach agent for guidance and support.

Troubleshooting

Issue: Hooks are not running.

Solution: Make sure the hooks are executable:

chmod +x .git/hooks/pre-commit
chmod +x .git/hooks/commit-msg

Issue: Tests are failing immediately.

Solution: Ensure you have at least one passing test:

npm test # Should see: Tests passed

Issue: Lint errors are blocking everything.

Solution: Enable auto-fix:

{
  "autoFix": true
}

Or fix manually:

npm run lint -- --fix

Integration

This skill integrates with Git repositories and various testing frameworks. It uses Git hooks to enforce TDD and ensure that all code changes are fully tested. The overnight-dev-coach agent provides guidance and support throughout the process.

# Example integration with Jest (Node.js)
{
  "testCommand": "npm test -- --coverage --watchAll=false",
  "lintCommand": "npm run lint",
  "autoFix": true
}
# Example integration with pytest (Python)
{
  "testCommand": "pytest --cov=. --cov-report=term-missing",
  "lintCommand": "flake8 . && black --check .",
  "autoFix": false
}

快速安装

/plugin add https://github.com/jeremylongshore/claude-code-plugins-plus/tree/main/overnight-dev

在 Claude Code 中复制并粘贴此命令以安装该技能

GitHub 仓库

jeremylongshore/claude-code-plugins-plus
路径: backups/skills-migration-20251108-070147/plugins/productivity/overnight-dev/skills/overnight-dev
aiautomationclaude-codedevopsmarketplacemcp

相关推荐技能

llamaguard

其他

LlamaGuard是Meta推出的7-8B参数内容审核模型,专门用于过滤LLM的输入和输出内容。它能检测六大安全风险类别(暴力/仇恨、性内容、武器、违禁品、自残、犯罪计划),准确率达94-95%。开发者可通过HuggingFace、vLLM或Sagemaker快速部署,并能与NeMo Guardrails集成实现自动化安全防护。

查看技能

sglang

SGLang是一个专为LLM设计的高性能推理框架,特别适用于需要结构化输出的场景。它通过RadixAttention前缀缓存技术,在处理JSON、正则表达式、工具调用等具有重复前缀的复杂工作流时,能实现极速生成。如果你正在构建智能体或多轮对话系统,并追求远超vLLM的推理性能,SGLang是理想选择。

查看技能

evaluating-llms-harness

测试

该Skill通过60+个学术基准测试(如MMLU、GSM8K等)评估大语言模型质量,适用于模型对比、学术研究及训练进度追踪。它支持HuggingFace、vLLM和API接口,被EleutherAI等行业领先机构广泛采用。开发者可通过简单命令行快速对模型进行多任务批量评估。

查看技能

langchain

LangChain是一个用于构建LLM应用程序的框架,支持智能体、链和RAG应用开发。它提供多模型提供商支持、500+工具集成、记忆管理和向量检索等核心功能。开发者可用它快速构建聊天机器人、问答系统和自主代理,适用于从原型验证到生产部署的全流程。

查看技能