# AGENTS.md - Mission Control Agent Swarm

## Project Overview

Mission Control is an ADHD-centric personal productivity hub built with Tauri 2.0 + React 19. It unifies GitLab activity (via `lore` CLI) and beads task tracking (via `br` CLI) into a single native interface.

**Core principle:** Surface THE ONE THING you should be doing right now.

## Architecture

```
Frontend (React 19 + Vite)     Backend (Tauri/Rust)
        |                              |
        |---- IPC (invoke) ----------->|
        |                              |
        |                     CLI Traits (mockable)
        |                              |
        |                     lore --robot (GitLab)
        |                     br (beads tasks)
        |                     bv --robot-* (triage)
```

## Key Files

| Path | Purpose |
|------|---------|
| `src/App.tsx` | Main React component |
| `src-tauri/src/lib.rs` | Tauri app setup |
| `src-tauri/src/data/lore.rs` | Lore CLI trait + types |
| `src-tauri/src/data/beads.rs` | Beads CLI trait + types |
| `src-tauri/src/commands/mod.rs` | Tauri IPC commands |
| `PLAN.md` | Full implementation plan |
| `CLAUDE.md` | Project-specific Claude instructions |

## Agent Coordination

### Team Name
`purrfect-beaming-lynx`

### Beads Task Management

Use `br` and `bv` for task coordination:

```bash
# Get triage recommendations
bv --robot-triage | jq '.triage.quick_ref.top_picks'

# List all beads
br list --json

# Claim a task
br update <bead-id> --status in_progress

# Complete a task
br close <bead-id> --reason "Completed: <summary>"

# Show bead details
br show <bead-id> --json
```

### Communication

Use SendMessage tool for agent-to-agent coordination:
- Message teammates directly by name
- Use broadcast sparingly (only for critical team-wide issues)
- Check task list after completing each task

### Current Status

**Completed:**
- Phase 0: Test infrastructure (Vitest, Playwright, Rust mocks)
- Phase 1: Tauri scaffold, React shell, CLI traits
- Phase 2: Bridge + Data Layer (GitLab-Beads sync, state persistence)
- Phase 3: UI Components (FocusView, QueueView, InboxView, Settings, Debug)
- Phase 4: Navigation + keyboard shortcuts (Cmd+1-4, Cmd+,)
- Phase 5: TanStack Query data layer, error boundaries
- bv triage integration (get_triage, get_next_pick commands)

**In Progress:**
- UI polish and edge case handling
- Decision log analysis features

## Development Commands

```bash
# Run dev (frontend + backend)
npm run tauri:dev

# Frontend only
npm run dev

# Run tests
npm run test           # Vitest
npm run test:e2e       # Playwright
cargo test             # Rust

# Lint
npm run lint
cargo clippy -- -D warnings
```

## Visual Verification

Playwright MCP tools can connect to `http://localhost:1420` for UI inspection:
- `browser_navigate` - Load pages
- `browser_snapshot` - Get accessibility tree
- `browser_take_screenshot` - Visual verification

## Critical Constraints

1. **CLI over libraries** - Shell out to `lore`/`br`/`bv`, don't import them
2. **Trait-based mocking** - All CLI interactions via traits for testability
3. **TDD** - Write failing test first, then implement
4. **No `any`** - Use `unknown` with type guards in TypeScript
5. **jj for VCS** - Use jj, not git (see global CLAUDE.md)