File Ownership Principles

Each file in the project is owned by exactly one agent. Agents must only modify files within their domain. This prevents conflicts, ensures accountability, and makes reviews straightforward.

General Rules

Domain	Owner	Examples
Business logic	Cursor	Hooks, services, API clients, utilities, logic-heavy components
Visual components	Lovable	Design system (`ui/`), layouts, navigation, styling, CSS
Config and docs	Claude Code	AGENTS.md, package.json, tsconfig, .ai/, docs/
Oversight files	Kimi Code	Reviews, instructions, status updates (via `.ai/`)

Typical File Ownership Map

code

src/
├── components/
│   ├── ui/              → Lovable (design system primitives)
│   └── [features]/      → Cursor (feature components with logic)
├── pages/               → Cursor (route pages with logic)
├── hooks/               → Cursor (custom hooks, except UI-only)
├── services/            → Cursor (API integration layer)
├── lib/                 → Cursor (shared utilities)
├── types/               → Claude Code (TypeScript type definitions)
├── App.tsx              → Claude Code (main router)
├── main.tsx             → Claude Code (app entry)
├── index.css            → Lovable (global styles)
└── App.css              → Lovable (app styles)

.ai/                     → Claude Code (coordination layer)
docs/                    → Claude Code (documentation)
package.json             → Claude Code (dependencies)
tsconfig*.json           → Claude Code (TypeScript config)
vite.config.ts           → Claude Code (build config)
tailwind.config.ts       → Claude Code (styling config)

For the complete and project-specific ownership map, always check .ai/boundaries.md.

Communication Folder Ownership

Folder	Who Writes	Who Reads
`.ai/tasks/`	Claude Code	All agents
`.ai/instructions/`	Kimi, Claude Code	Target agent
`.ai/reviews/`	Kimi, Claude Code	Submitting agent + Human
`.ai/reports/`	All agents	Human + all agents
`.ai/chats/`	All agents	All agents
`.ai/ideas/`	Claude Code	All agents + Human
`.ai/templates/`	Claude Code	All agents

How to Check Boundaries Before Modifying Files

Before Starting Work

•Read the task brief — .ai/tasks/TASK-XXX.md lists files in scope
•Check AGENTS.md — The "Owns" section under your agent role
•Check .ai/boundaries.md — For the complete ownership map
•When in doubt, ask — Use .ai/chats/ to clarify with the orchestrator

Before Committing

•Review your diff — git diff --stat to see all modified files
•Verify each file — Is every modified file in your domain?
•Remove unauthorized changes — git checkout -- <file> for files outside your domain
•Document exceptions — If you must touch another agent's file, explain in the commit body

Boundary Violation Consequences

Severity	Violation	Consequence
High	Modifying another agent's core files	REJECT — must revert changes
Medium	Modifying shared config without approval	CHANGES REQUESTED — remove or get approval
Low	Minor touch to adjacent file (e.g., import fix)	Note in review — approve if justified

Examples of Violations

Agent	Violation	Why It's Wrong
Cursor edits `src/components/ui/Button.tsx`	Lovable owns design system	Cursor should request Lovable to make UI changes
Lovable adds a hook to `src/hooks/use-auth.ts`	Cursor owns hooks with logic	Lovable should request Cursor to add the hook
Any agent edits `package.json`	Claude Code owns config	Request Claude Code to add dependencies
Any agent edits `.ai/tasks/`	Claude Code owns task briefs	Report status via `.ai/reports/` instead

Exceptions

Some situations justify cross-boundary changes:

•Import statements — Adding an import for a new export is usually acceptable
•Type definitions — Updating shared types may be necessary
•Emergency fixes — Critical bugs may require immediate cross-boundary fixes

All exceptions must be documented in the commit message body and flagged in the review.

Reference

•Complete ownership map: .ai/boundaries.md
•Agent roles and domains: AGENTS.md
•Review process: review-checklist skill