AI Workflow System

Structured workflow for AI-assisted development. Prevents chaos, maintains discipline.

For detailed context on the workflow system, validation evidence, and comparison to traditional artifacts, see What Is This Repo?

---

Three Modes (Don't Mix)

Exploration → Are we building the right thing?

• Question architecture, make decisions
• Output: What to build

Planning → How do we build it?

• Break into atomic tasks with TDD
• Output: Implementation roadmap

Implementation → Build it

• Test-first workflow, commit per phase
• Output: Working, tested code

---

Quick Start

For Humans:

1. Create brief: docs/workflow/brief-creation.md
2. Work with AI through modes
3. Brief tracks progress

For AI:

1. Read docs/workflow/session-start.md first
2. Follow mode-specific workflow
3. Update brief as you go

---

File Structure

docs/
  ├── workflow/        # Process docs (AI reads these)
  ├── briefs/          # Work packages (active/completed)
  ├── context/         # Mission, architectural context
  ├── standards/       # Code/process standards
  └── TAG_INDEX.json   # Doc navigation

---

Key Principle

Mode discipline prevents context collapse. Don't code during exploration. Don't debate architecture during implementation.

---

Customization

Adapt: Constraints, tags, project specifics Keep: Core mode structure, TDD workflow, phase boundaries

See change protocol in workflow docs before modifying process.