feat: Unified Inbox — cross-session notification center

package.json

@@ -244,6 +244,7 @@
     "react-diff-view": "^3.3.2",
     "react-markdown": "^10.1.0",
     "react-syntax-highlighter": "^16.1.0",
+    "react-window": "^2.2.7",
     "reactflow": "^11.11.4",
     "recharts": "^3.6.0",
     "rehype-raw": "^7.0.0",

playbooks/agent-inbox/LOOP_00001_COVERAGE_REPORT.md

@@ -0,0 +1,167 @@
+---
+type: report
+title: Agent Inbox — Baseline Test Coverage Report
+created: 2026-02-15
+tags:
+  - test-coverage
+  - agent-inbox
+  - phase-07a
+related:
+  - "[[UNIFIED-INBOX-07a]]"
+  - "[[UNIFIED-INBOX-07b]]"
+---
+
+# Agent Inbox — Baseline Test Coverage Report
+
+> **Measured:** 2026-02-15 | **Branch:** `feature/unified-inbox` | **Framework:** Vitest + V8
+
+---
+
+## Overall Line Coverage: 60.09%
+
+| Metric       | Covered | Total  | Percentage |
+|-------------|---------|--------|-----------|
+| **Statements** | 35,138  | 59,286 | **59.26%** |
+| **Branches**   | 22,947  | 42,470 | **54.03%** |
+| **Functions**  | 7,314   | 12,645 | **57.84%** |
+| **Lines**      | 33,387  | 55,560 | **60.09%** |
+
+- **Target:** 80%
+- **Gap to Target:** ~20 percentage points
+- **Test Suite:** 452 test files, 19,336 tests passing, 107 skipped, 0 failures
+
+---
+
+## Agent Inbox Feature — Coverage Breakdown
+
+| File | Stmts | Branch | Funcs | Lines | Tests | Status |
+|------|-------|--------|-------|-------|-------|--------|
+| `AgentInbox.tsx` | 87.57% | 82.69% | 86.84% | 91.60% | 88 | Above target |
+| `useAgentInbox.ts` | 98.68% | 95.65% | 100% | 98.38% | 38 | Excellent |
+| `agent-inbox.ts` (types) | 100% | 100% | 100% | 100% | — | Complete |
+| `agentInboxHelpers` (cross-file) | — | — | — | — | 17 | Pure functions |
+| `modalStore.ts` (shared) | 69.45% | 51.27% | 40% | 70.22% | 66 | Below target |
+
+**Agent Inbox Total Tests:** 143+ across 4 test files (88 + 38 + 17 in dedicated files)
+
+### Key Observations
+
+- `useAgentInbox.ts` is **near-perfect** at 98.38% line coverage — only line 25 (an early guard) uncovered.
+- `AgentInbox.tsx` is **well-covered** at 91.60% lines — small gaps in lines 448-452 and 638-641.
+- `modalStore.ts` is **below target** at 70.22% — Agent Inbox modal functions (lines 628-736) are partially uncovered. This is shared infrastructure, not Agent Inbox-specific.
+
+---
+
+## Coverage by Module (Selected Highlights)
+
+### Well-Covered Modules (>80% lines)
+
+| Module | Lines | Notes |
+|--------|-------|-------|
+| `shared/` | 92.66% | Shared utilities, formatters |
+| `renderer/stores/` | 89.76% | State management stores |
+| `web/components/` | 97.55% | Web UI components |
+| `web/hooks/` | 97.79% | Web hooks |
+| `web/utils/` | 99.54% | Web utilities |
+| `renderer/hooks/batch/` | 84.57% | Batch processing hooks |
+| `renderer/hooks/remote/` | 86.70% | SSH/remote hooks |
+
+### Below-Target Modules (<80% lines)
+
+| Module | Lines | Notes |
+|--------|-------|-------|
+| `renderer/hooks/session/` | 31.56% | Session management hooks (largest gap) |
+| `renderer/hooks/props/` | 0% | Panel prop hooks (no tests at all) |
+| `renderer/hooks/symphony/` | 43.44% | Symphony/contribution hooks |
+| `renderer/hooks/keyboard/` | 63.19% | Keyboard handler (complex, large) |
+| `renderer/hooks/input/` | 69.80% | Input processing hooks |
+| `renderer/hooks/git/` | 69.88% | Git management hooks |
+| `renderer/hooks/ui/` | 65.53% | UI utility hooks |
+| `renderer/services/` | 65.69% | Services layer |
+| `renderer/utils/` | 77.92% | Renderer utilities |
+| `web/` (App) | 26.53% | Web app entry point |
+
+---
+
+## Lowest Coverage Files (Critical Gaps)
+
+| File | Lines | Category |
+|------|-------|----------|
+| `hooks/session/useSessionUpdates.ts` | 1.04% | Session state management |
+| `hooks/session/useRealtimeTracker.ts` | 3.94% | Session realtime tracking |
+| `hooks/session/useSessionNavigation.ts` | 0% | Session navigation |
+| `hooks/session/usePinnedSessions.ts` | 0% | Pinned sessions |
+| `hooks/props/useMainPanelProps.ts` | 0% | Main panel props |
+| `hooks/props/useRightPanelProps.ts` | 0% | Right panel props |
+| `hooks/props/useSessionListProps.ts` | 0% | Session list props |
+| `hooks/ui/useAppHandlers.ts` | 0% | App-level handlers |
+| `hooks/ui/useThemeStyles.ts` | 0% | Theme style hook |
+| `hooks/input/useInputSync.ts` | 0% | Input synchronization |
+| `hooks/agent/useSendAndContinue.ts` | 0% | Agent send & continue |
+| `hooks/symphony/useContribution.ts` | 0% | Symphony contributions |
+| `hooks/symphony/useContributorStats.ts` | 0% | Contributor stats |
+| `services/contextGroomer.ts` | 22.44% | Context grooming service |
+| `utils/remarkSmartFormatTable.ts` | 5.55% | Table formatting utility |
+
+---
+
+## Existing Test Patterns
+
+The project follows consistent patterns across all 452 test files:
+
+1. **Testing Stack:** Vitest + @testing-library/react + jsdom
+2. **Hook Testing:** `renderHook` with `act()` for state updates
+3. **Component Testing:** `render` + `screen` queries + `fireEvent`
+4. **Mock Strategy:** `vi.mock()` for modules, `vi.fn()` for functions
+5. **Factory Functions:** `makeSession()`, `makeTab()`, etc. for test data creation
+6. **Semantic Queries:** `getByRole`, `getByText`, `getAllByRole` preferred over test IDs
+7. **Accessibility Testing:** ARIA attributes validated in dedicated test blocks
+8. **Setup:** Global `setup.ts` with jsdom environment, tab-indented files
+
+---
+
+## Recommendations
+
+### Quick Wins (High coverage gain, low effort)
+
+These modules have partial coverage and could reach 80%+ with targeted tests:
+
+| Target | Current | Effort | Notes |
+|--------|---------|--------|-------|
+| `modalStore.ts` Agent Inbox lines | 70% | Low | Add tests for inbox-related modal actions (lines 628-736) |
+| `renderer/utils/` assorted | 78% | Low | Many files at 80-95%, a few at 0% pulling average down |
+| `renderer/hooks/settings/` | 76% | Medium | Large file with many branches |
+| `AgentInbox.tsx` remaining lines | 92% | Low | Just lines 448-452, 638-641 uncovered |
+
+### Requires Setup (Medium effort, infrastructure needed)
+
+| Target | Current | Effort | Blocker |
+|--------|---------|--------|---------|
+| `hooks/session/` cluster | 32% | High | Complex state management, needs extensive mocking |
+| `hooks/keyboard/` handler | 63% | Medium | Large file (750+ lines), needs keyboard event simulation |
+| `hooks/input/` processing | 70% | Medium | Complex input pipeline with debouncing |
+| `services/contextGroomer.ts` | 22% | High | Requires IPC mocking for main process communication |
+
+### Skip for Now (Low ROI or non-essential)
+
+| Target | Reason |
+|--------|--------|
+| `hooks/props/*` (0%) | Pure prop-passing hooks, low logic density |
+| `web/App.tsx` (27%) | Web app entry point, hard to unit test |
+| `utils/confetti.ts` (0%) | Animation utility, visual-only |
+| `utils/clipboard.ts` (0%) | Browser API wrapper, hard to test in jsdom |
+| `utils/formatters.ts` (0%) | Renderer-side formatters may be redundant with shared/formatters |
+| Type definition files (0%) | No executable code to test |
+
+---
+
+## Agent Inbox Specific — Next Steps
+
+The Agent Inbox feature is in **strong shape** at 91-98% coverage for its core files. To reach 80% project-wide target, the effort should focus on:
+
+1. **Close the `modalStore.ts` gap** — Add tests for the Agent Inbox modal state transitions (lines 628-736)
+2. **Cover `AgentInbox.tsx` edge cases** — Lines 448-452 and 638-641 are small gaps
+3. **Integration test** — A single test exercising the full pipeline (store → hook → component → user interaction)
+4. **Performance test** — Verify rendering with 100+ inbox items doesn't regress
+
+The broader project coverage gap (60% vs 80% target) is overwhelmingly in non-Agent-Inbox modules, particularly the session management and keyboard handling hooks.

playbooks/agent-inbox/LOOP_00001_GAPS.md

@@ -0,0 +1,191 @@
+---
+type: analysis
+title: Agent Inbox — Test Coverage Gaps
+created: 2026-02-15
+tags:
+  - test-coverage
+  - agent-inbox
+  - phase-07b
+related:
+  - "[[UNIFIED-INBOX-07a]]"
+  - "[[UNIFIED-INBOX-07b]]"
+  - "[[LOOP_00001_COVERAGE_REPORT]]"
+---
+
+# Agent Inbox — Test Coverage Gaps
+
+> **Measured:** 2026-02-15 | **Branch:** `feature/unified-inbox` | **Source:** Phase 07a coverage report + manual code inspection
+
+---
+
+## Summary
+
+The Agent Inbox feature is in **strong shape** (91–98% line coverage for core files). This document catalogs the remaining small gaps to close in Phase 07c/07d.
+
+| File | Current Lines | Target | Gap Size |
+|------|--------------|--------|----------|
+| `AgentInbox.tsx` | 91.60% | 95%+ | ~8 lines |
+| `useAgentInbox.ts` | 98.38% | 100% | 1 line |
+| `modalStore.ts` (shared) | 70.22% | 80% | ~2 lines (Agent Inbox specific) |
+| `agent-inbox.ts` (types) | 100% | 100% | None |
+
+**Total new tests estimated:** 8–12 test cases
+
+---
+
+## Gap 1: `AgentInbox.tsx` — `findRowIndexForItem` fallback return
+
+| Field | Value |
+|-------|-------|
+| **File** | `src/renderer/components/AgentInbox.tsx` |
+| **Function** | `findRowIndexForItem` (lines 446–455) |
+| **Uncovered Lines** | 448–452 |
+| **Type** | Edge Case |
+| **Current Coverage** | Partial — only called implicitly via `scrollToRow` effect |
+| **Why It Matters** | The for-loop body (line 450 match check) and fallback `return 0` (line 452) are not directly tested. If the loop logic breaks, items won't scroll into view correctly. |
+| **Description** | This callback maps an item index to a row index (accounting for group headers in grouped mode). The fallback `return 0` handles the case where no matching row is found — e.g., when selectedIndex is out of bounds or rows are empty. |
+| **Suggested Test Approach** | **Unit test** — Extract `buildRows` (already module-scoped) and test `findRowIndexForItem` logic indirectly by: (1) rendering in grouped mode and navigating to verify scroll behavior, OR (2) testing `buildRows` directly and verifying row indices match expected positions. A simpler approach: test that selecting an item in grouped mode with headers produces correct `aria-activedescendant`. |
+
+---
+
+## Gap 2: `AgentInbox.tsx` — Close button hover handlers
+
+| Field | Value |
+|-------|-------|
+| **File** | `src/renderer/components/AgentInbox.tsx` |
+| **Function** | Close button `onMouseEnter`/`onMouseLeave` (lines 637–641) |
+| **Uncovered Lines** | 638–641 |
+| **Type** | Edge Case (Visual) |
+| **Current Coverage** | 0% — no test fires mouseenter/mouseleave on close button |
+| **Why It Matters** | Low risk — these are hover style handlers. However, they do modify `style.backgroundColor` imperatively, which could cause visual regressions if removed or broken. |
+| **Description** | `onMouseEnter` sets the close button background to `${theme.colors.accent}20` (accent at 12.5% opacity). `onMouseLeave` resets it to `'transparent'`. |
+| **Suggested Test Approach** | **Component test** — Fire `mouseEnter` and `mouseLeave` events on the close button (`screen.getByTitle('Close (Esc)')`) and assert `style.backgroundColor` changes. Low effort, 2 test cases. |
+
+---
+
+## Gap 3: `useAgentInbox.ts` — `matchesFilter` default branch
+
+| Field | Value |
+|-------|-------|
+| **File** | `src/renderer/hooks/useAgentInbox.ts` |
+| **Function** | `matchesFilter` (lines 12–27) |
+| **Uncovered Lines** | 25 (`default: return false`) |
+| **Type** | Edge Case (Defensive Guard) |
+| **Current Coverage** | 0% for this specific line — all 3 valid filter modes are tested, but the `default` branch (unreachable with current types) is not |
+| **Why It Matters** | Very low risk — TypeScript enforces `InboxFilterMode` is `'all' | 'needs_input' | 'ready'`, so this line is unreachable at compile time. It exists as a defensive guard. |
+| **Description** | The `default` case returns `false` for any unrecognized filter mode. Since `InboxFilterMode` is a union type, this can only be reached via type casting (e.g., `'invalid' as InboxFilterMode`). |
+| **Suggested Test Approach** | **Unit test** — Call `useAgentInbox` with an invalid filter mode cast via `as InboxFilterMode` and verify it returns an empty array. Alternatively, accept this as an intentional uncovered guard line (1 line = 0.02% impact). |
+
+---
+
+## Gap 4: `modalStore.ts` — `setAgentInboxOpen` action
+
+| Field | Value |
+|-------|-------|
+| **File** | `src/renderer/stores/modalStore.ts` |
+| **Function** | `setAgentInboxOpen` (lines 527–529) |
+| **Uncovered Lines** | 528–529 (open/close branches) |
+| **Type** | Unit |
+| **Current Coverage** | 0% — no test in `modalStore.test.ts` covers the Agent Inbox modal action |
+| **Why It Matters** | Medium risk — if the modal type string `'agentInbox'` is misspelled or the action is removed, the Inbox won't open/close. This is the bridge between the keyboard shortcut and the modal rendering. |
+| **Description** | `setAgentInboxOpen(true)` calls `openModal('agentInbox')`, `setAgentInboxOpen(false)` calls `closeModal('agentInbox')`. These are thin wrappers around the generic modal store machinery. |
+| **Suggested Test Approach** | **Unit test** — In `modalStore.test.ts`, add a test block: call `setAgentInboxOpen(true)`, assert `isOpen('agentInbox')` is true. Call `setAgentInboxOpen(false)`, assert it's false. 2 test cases, very low effort. |
+
+---
+
+## Gap 5: `AgentInbox.tsx` — Focus/blur outline handlers
+
+| Field | Value |
+|-------|-------|
+| **File** | `src/renderer/components/AgentInbox.tsx` |
+| **Function** | `onFocus`/`onBlur` handlers on InboxItemCardContent (lines 120–126) and SegmentedControl buttons (lines 285–291) |
+| **Uncovered Lines** | Not in the 07a report as uncovered, but no test explicitly verifies focus ring behavior |
+| **Type** | Edge Case (Accessibility) |
+| **Current Coverage** | Partial — focus is tested for Tab cycling but not the visual outline style |
+| **Why It Matters** | Medium risk for accessibility — focus rings are critical for keyboard users. If `outline` styling breaks, keyboard navigation becomes invisible. |
+| **Description** | `onFocus` sets `outline: 2px solid ${theme.colors.accent}` with `-2px` offset. `onBlur` removes it. Applied to both item cards and segmented control buttons. |
+| **Suggested Test Approach** | **Component test** — Focus an item card via `fireEvent.focus()` and check `style.outline`. Then blur and verify outline is removed. 2–3 test cases. |
+
+---
+
+## Gap 6: `AgentInbox.tsx` — `InboxRow` null guard for missing row
+
+| Field | Value |
+|-------|-------|
+| **File** | `src/renderer/components/AgentInbox.tsx` |
+| **Function** | `InboxRow` (lines 310–363) |
+| **Uncovered Lines** | 323 (`if (!row) return null`) |
+| **Type** | Edge Case (Defensive Guard) |
+| **Current Coverage** | 0% for null guard — react-window always passes valid indices |
+| **Why It Matters** | Very low risk — this guard protects against react-window passing an out-of-bounds index, which shouldn't happen in practice. |
+| **Description** | If `rows[index]` is undefined (e.g., due to a race condition between rowCount and rows array), the component returns null instead of crashing. |
+| **Suggested Test Approach** | Skip — this is a defensive guard that cannot be triggered through normal UI interaction. Testing it requires directly rendering `InboxRow` with an invalid index, which tests implementation details rather than behavior. Accept as intentional uncovered guard. |
+
+---
+
+## Untested Branches Summary
+
+| Location | Branch Type | Tested Path | Untested Path |
+|----------|-------------|-------------|---------------|
+| `matchesFilter` switch | switch/default | `all`, `needs_input`, `ready` | `default` (unreachable) |
+| `findRowIndexForItem` for-loop | loop exit | N/A (implicit via scroll) | Loop body + fallback return 0 |
+| Close button hover | if/else (enter/leave) | Neither | Both `onMouseEnter` and `onMouseLeave` |
+| `InboxRow` null guard | if/early-return | Normal row rendering | `!row` guard |
+| `setAgentInboxOpen` | boolean branch | Neither | Both `true` and `false` |
+
+---
+
+## Untested Error Handling
+
+| Location | Error Type | Status |
+|----------|-----------|--------|
+| `useAgentInbox` — undefined `aiTabs` | null guard (`?? []`) | **Tested** (line 93 of hook test) |
+| `useAgentInbox` — undefined `logs` | null guard (`?? []`) | **Tested** (line 330 of hook test) |
+| `useAgentInbox` — null log text | null guard (`?.text`) | **Tested** (line 549 of hook test) |
+| `useAgentInbox` — empty session id | falsy guard (`!session.id`) | **Tested** (line 103 of hook test) |
+| `deriveTimestamp` — invalid timestamp | fallback chain | **Tested** (line 610 of hook test) |
+| `resolveStatusColor` — unknown color key | fallback (`?? textDim`) | **Not tested** — would need a session state not in `STATUS_COLORS` map |
+
+---
+
+## Untested Edge Cases
+
+| Location | Edge Case | Priority |
+|----------|-----------|----------|
+| `AgentInbox.tsx` — context bar at exactly 0% | Width `0%` rendering | Low |
+| `AgentInbox.tsx` — negative contextUsage | `Math.max(value, 0)` clamp | Low |
+| `buildRows` — empty items array in grouped mode | Returns empty array | Low |
+| `AgentInbox.tsx` — `selectedItemId` when selectedIndex > items.length | Returns `undefined` | Low |
+| `AgentInbox.tsx` — `listHeight` on server (typeof window === 'undefined') | Returns 400 fallback | Low |
+
+---
+
+## Recommended Test Priority
+
+### Must Have (High ROI, Low Effort)
+
+1. **Gap 4: `setAgentInboxOpen` in modalStore** — 2 tests, ensures modal open/close works
+2. **Gap 2: Close button hover** — 2 tests, covers the last uncovered lines in AgentInbox.tsx
+
+### Nice to Have (Medium ROI)
+
+3. **Gap 1: `findRowIndexForItem` fallback** — 1–2 tests, validates grouped mode scroll behavior
+4. **Gap 5: Focus ring handlers** — 2–3 tests, accessibility assurance
+
+### Skip (Low ROI)
+
+5. **Gap 3: `matchesFilter` default** — Unreachable via TypeScript types
+6. **Gap 6: `InboxRow` null guard** — Defensive guard, never triggered in practice
+
+---
+
+## Projected Coverage After Closing Gaps
+
+| File | Current | After Gaps 1–4 | After All |
+|------|---------|----------------|-----------|
+| `AgentInbox.tsx` | 91.60% | ~95% | ~97% |
+| `useAgentInbox.ts` | 98.38% | 98.38% | ~99.5% |
+| `modalStore.ts` | 70.22% | ~71% | ~71% |
+| **Agent Inbox Overall** | ~93% | ~95% | ~97% |
+
+> Note: `modalStore.ts` coverage gains from Agent Inbox tests are minimal (~0.5%) since the file is 737 lines and Agent Inbox occupies only 3 lines. The broader modalStore coverage gap is a separate concern.