docs: establish project documentation and AI context

Start with the requested docs/ directory structure and root AGENTS.md deliverables. Use the acceptance criteria to check that each architectural decision has a documented home, current and planned behavior are distinguished, and no document promises unimplemented functionality.

Objective

Create the initial docs/ system for QueueIt so both humans and coding agents have a single source of truth before feature work grows.

Problem

The project should not repeat the documentation and architecture drift that caused problems in previous projects. AI agents must be able to understand the product, constraints, architecture, and current decisions without guessing.

Deliverables

Create:

docs/
├── README.md
├── 00-product/
│   ├── vision.md
│   ├── scope.md
│   └── glossary.md
├── 01-architecture/
│   ├── overview.md
│   ├── extension-runtime.md
│   ├── queue-engine.md
│   └── adapter-system.md
├── 02-chatgpt/
│   ├── integration.md
│   ├── message-lifecycle.md
│   └── completion-detection.md
├── 03-ui/
│   ├── principles.md
│   └── screens.md
├── 04-data/
│   ├── state-model.md
│   └── persistence.md
├── 05-security/
│   └── threat-model.md
├── 06-monetization/
│   ├── free-paid.md
│   └── entitlement-model.md
├── 07-testing/
│   ├── strategy.md
│   └── e2e.md
└── 08-roadmap/
    └── milestones.md

Also add a root-level AGENTS.md describing how AI agents must work on QueueIt, including reading the docs first, preserving architectural boundaries, avoiding speculative dependencies, updating docs when decisions change, and running validation before claiming completion.

Planned work

• #2 Product requirements and scope
• #3 Extension architecture
• #4 Queue state machine and persistence
• #5 ChatGPT lifecycle and completion detection
• #6 UI/UX system
• #7 Free/paid product model
• #8 Security, privacy, and permissions
• #9 Testing and reliability
• #10 Development workflow and AI-agent rules

Acceptance criteria

• Every architectural decision has a documented home.
• A new AI coding agent can read the docs and understand how QueueIt works.
• Docs distinguish current implementation from planned behavior.
• No document promises functionality that has not been implemented.
• The documentation structure remains small and navigable.