AI Workflow Requirements Notes
A workflow foundation where AI agents and humans collaborate on development. Exploring a design that balances automation and quality assurance.
Role
Personal project
Tech Stack
TypeScript, pnpm, OpenTelemetry, CI/CD, Markdown
Concept
A workflow foundation where AI agents (Claude, etc.) and human developers collaborate to deliver efficient, high-quality development. Through text-based design management and automated quality checks, the goal is to speed up delivery while avoiding knowledge silos.
In-Progress Considerations
- Standardizing agent instructions: Use AGENTS.md and claude.md to clarify AI behavior rules and keep output quality consistent.
- Automated quality checks: Integrate lint/test/security checks into the CI/CD pipeline to reduce human error.
- Observability: Visualize development processes with OpenTelemetry and automate bottleneck detection.
- Documentation-driven development: Build a workflow centered on OVERVIEW.md to minimize drift between specs and implementation.
Notes
As a personal project, I am exploring the best process for collaborating with AI. By combining pnpm-based dependency management and Markdown-driven documentation, I maintain a lightweight, reproducible environment while applying a Human-in-the-Loop mindset to balance quality and automation.