overnight-dev
Run Claude autonomously for 6-8 hours overnight using Git hooks that enforce TDD - wake up to fully tested features
Installation
Open Claude Code and run this command:
/plugin install overnight-dev@claude-code-plugins-plus
Use --global to install for all projects, or --project for current project only.
What It Does
Go to bed. Wake up to fully tested features.
Run Claude autonomously for 6-8 hours overnight using Git hooks that enforce test-driven development. No more "should work" - only "does work."
Features
Test-Driven Development Enforcement
Git hooks force TDD:
- Write tests first
- Implementation after
- Commit only when green
- No shortcuts, no broken code
Autonomous Debugging
When tests fail:
- Claude reads the error
- Analyzes the problem
- Forms a hypothesis
- Makes a fix
- Tries to commit again
- Repeat until green
You never have to intervene. Just check progress in the morning.
Progress Tracking
Watch overnight sessions in real-time:
cat .overnight-dev-log.txt
[22:15] Session started: JWT Authentication
[22:20] Tests: 12/12 passing
[23:10] Tests: 18/18 passing - Auth routes done
[00:30] Tests: 24/24 passing - Middleware complete
[02:15] Tests: 35/35 passing - Integration tests done
[04:00] Tests: 47/47 passing - Coverage 94%
[06:45] SESSION COMPLETE
️ Flexible Configuration
Works with any test framework:
Node.js:
{
"testCommand": "npm test",
"lintCommand": "npm run lint"
}
Python:
{
"testCommand": "pytest --cov=.",
"lintCommand": "flake8 ."
}
Rust, Go, PHP, Ruby - All supported!
Smart Agent Guidance
Includes overnight-dev-coach agent:
- Guides you through setup
- Plans overnight tasks
- Debugs failing tests
- Tracks progress
- Celebrates success
Activate by mentioning "overnight development" or asking about autonomous coding.
Skills (1)
Automates software development overnight using git hooks to enforce test-driven Use when appropriate context detected.
How It Works
Starting an Overnight Session
- Define a clear goal:
Task: Build payment integration with Stripe
Success: All tests pass, 85%+ coverage, fully documented
- Start coding:
- Write tests first (TDD)
- Implement features
- Try to commit
- Let hooks guide you:
- Tests failing? Hooks block the commit
- Claude sees the error → debugs → fixes → retries
- Tests passing? Commit succeeds → Continue
- Go to sleep:
- Claude keeps iterating
- Hooks enforce quality
- No broken code gets committed
- Wake up to success:
- All tests passing
- Features complete
- Clean git history
Good Overnight Tasks
"Build user authentication with JWT (90% coverage)"
- Clear goal
- Testable
- Well-defined scope
"Add payment processing with Stripe integration"
- Specific feature
- Integration tests possible
- Success criteria clear
"Refactor database layer to use repository pattern"
- Existing tests ensure no regression
- Clear before/after state
Bad Overnight Tasks
"Make the app better"
- Too vague
- No clear success criteria
- Can't be tested
"Design the perfect UI"
- Subjective
- Hard to test
- Requires human judgment
"Research best practices"
- No code output
- No tests to enforce
- Not autonomous-friendly
FAQ
A: Yes. Hooks enforce tests, Claude debugs failures, morning brings green tests.
Q: What if tests fail all night?
A: Set maxAttempts to alert you. But in practice, Claude usually gets them green within a few tries.
Q: Can I use this for non-JS projects?
A: Absolutely! Works with any language that has tests and linting.
Q: How much does this cost?
A: Depends on your Claude usage. Overnight sessions typically use 1-3 hours of Claude time.
Q: Will this replace me?
A: No. You define the goals, write high-level tests, and make architectural decisions. Claude handles the tedious implementation and debugging.
Q: What if I don't have tests?
A: Write at least one test first. Hooks need something to run.