Coding Prompt Bank
Core philosophy and tier-based requirements for AI coding.
Philosophy
code
CORRECTNESS > SPEED Working first attempt beats fast iteration + debug. Mental simulation catches 80% of bugs before execution. EVERY LINE IS LIABILITY Config > code. Native > custom. Existing > new. Delete what doesn't earn its place. PARSE, DON'T READ Grep. Find. Read signatures first. Never ask for overview—discover it.
Project Tiers
Tier 1: Hackathon/Prototype
- •Speed matters most
- •Skip tests unless critical path
- •Minimal documentation
- •Technical debt acceptable
Tier 2: Default/MVP (most projects)
- •Balance speed and quality
- •Core path tests required
- •Basic documentation
- •Reasonable architecture
Tier 3: Production/Enterprise
- •Quality over speed
- •Comprehensive tests (80%+ coverage)
- •Full documentation
- •Robust error handling
- •Security review required
Universal Rules
- •Investigate before implement - Search for existing patterns
- •Type everything - TypeScript strict, Python type hints
- •Config over code - Magic numbers in config files
- •Fail fast - Validate early, clear error messages
- •Atomic commits - One logical change per commit
Default Behaviors
When Implementing
- •Research existing solutions first
- •Define interfaces before implementation
- •Write tests for core logic
- •Document public APIs
When Debugging
- •Reproduce first
- •Isolate the problem
- •Fix root cause, not symptom
- •Add regression test
Before Completing
- •Run type checker
- •Run linter
- •Run tests
- •Review changes
Integration
Loaded automatically on project initialization. Tier detected from:
- •
package.jsonkeywords - •README indicators
- •User specification