context-driven-development

Context-driven development treats project context as a managed artifact stored alongside code. It uses structured documents like product.md, tech-stack.md, and workflow.md to give AI assistants and team members a persistent single source of truth.

Run the setup command with existing codebase detection for brownfield projects. The system analyzes your code, configs, and documentation to pre-populate artifacts, which you then review and refine to match desired patterns.

Update product.md when features complete, vision changes, or target users shift. Update tech-stack.md before adding dependencies, upgrading major versions, or changing infrastructure, documenting the rationale and version constraints.

Common anti-patterns include stale context, information sprawl across locations, implicit knowledge never written down, and over-specification. Fix these by updating context with each track and keeping artifacts focused on decisions affecting behavior.

Yes, context persistence supports multi-session development. Start sessions by reading index.md and tracks.md, and end them by updating plan.md with progress, blockers, and status changes so the next session resumes cleanly.