LeanAI
Open Source

AI Code with Guardrails

The bottleneck has moved from code generation to code comprehension. Our open-source Claude Bootstrap library ensures every line of AI-generated code is simple, secure, and verifiable.

The Problem with Unconstrained AI Code

AI can generate thousands of lines of code in seconds. But without guardrails, you get:

  • ✗Sprawling functions that nobody can review
  • ✗Hardcoded secrets and security vulnerabilities
  • ✗No tests, no confidence in correctness
  • ✗Giant commits that are impossible to debug
# Without guardrails
function doEverything() {
// 500 lines of spaghetti
// hardcoded API keys
// zero tests
// good luck debugging
}

Our Quality Guardrails

Battle-tested constraints that keep AI-generated code maintainable

[ TEST ]

TDD-First Enforcement

Every feature starts with a failing test. No implementation without red-green validation. 80% minimum coverage enforced.

[ SIMPLE ]

Complexity Constraints

20-line max per function. 200-line max per file. 3 parameters max. 2 levels of nesting. Simple code is maintainable code.

[ SECURE ]

Security-First

No secrets in code. Pre-commit security scanning. Dependency auditing. Centralized credential management.

[ ATOMIC ]

Atomic Commits

Green: ≤5 files, ≤200 lines. Yellow: 6-10 files. Red: >10 files. Every commit is reviewable and reversible.

Automatic TDD Loop

Our "Ralph Wiggum" system automatically transforms natural language requests into TDD loops that iterate until tests pass—without manual intervention.

1
Red
Write failing test first
2
Green
Write minimal code to pass
3
Refactor
Clean up while tests stay green
↺
Iterate
Loop continues until all tests pass
# Claude Bootstrap TDD Loop
$ Add user authentication
✗ auth.test.ts - 3 failing
→ Writing implementation...
○ auth.test.ts - 1 failing
→ Fixing edge case...
✓ auth.test.ts - all passing
✓ Coverage: 94%
Done! 3 files, 47 lines, 12 tests

Every Commit is Verified

Pre-commit hooks and CI/CD pipelines catch issues before they reach production

Linting
ESLint/Prettier with auto-fixes
Type Safety
Strict TypeScript checking
Test Coverage
80% minimum enforced
Security Scan
Trufflehog secret detection
Dependency Audit
Known vulnerability checks
Build Verification
Production build must pass

36+ Battle-Tested Skills

Markdown-based patterns that encode our best practices for every technology we use

Core

TDDSecuritySession ManagementProject Tooling

Languages

TypeScriptPythonShell

Frameworks

ReactNext.jsReact NativeFastAPIExpress

Databases

SupabasePostgreSQLDrizzle ORM

Testing

VitestPlaywrightE2E Testing

AI Patterns

LLM IntegrationAgentic WorkflowsRAG

Quality is Non-Negotiable

Every project we build uses Claude Bootstrap and our quality guardrails. Your AI MVP will be fast, but it will also be maintainable, secure, and tested.