Guided learning project using the project-lab workflow. This is NOT a typical codebase - it exists to teach Node.js internals through hands-on implementation.
Code is split between AI and Human. Every commit message MUST document which files/lines were written by whom.
- Project scaffolding, config, directory structure
- Type definitions, interfaces, enums
- Function signatures with doc comments
- Glue code connecting modules
- Test cases (after human implements the logic)
- Helper utilities (logging, error types, config parsing)
- Core algorithms (parsers, state machines)
- Design decisions with multiple valid approaches
- Performance-critical code paths
- Protocol implementations
- The "interesting" conditional logic
Step N: [description]
## Authorship
### AI-generated:
- [file]: [what]
### HUMAN-written:
- [file] lines X-Y: [what they implemented]
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Exactly ONE
TODO(human)block per implementation step - Surrounding code must compile with the TODO replaced by a stub (throw/panic)
- Include input/output contract in comments above the TODO
- Do NOT write the TODO(human) logic for the user unless they explicitly ask (push back once first)
- Bun runtime, TypeScript strict mode, vitest
node:netfor raw TCP - nohttpmodule, no Express