feat(desktop): add monitor scheduler, Playwright CDP layer, and account health

Move monitor-task scheduling authority onto the client with a durable
file-backed queue that survives restart, drops stale cross-day tasks,
enforces per-platform serialism, and adapts global concurrency from
Electron process metrics. Publish tasks keep their existing FIFO.

Add a hidden Playwright CDP manager that attaches to Electron Chromium
on account session partitions, lets adapters opt into `executionMode:
"playwright"`, and leaves the existing hidden WebContentsView path in
place for current adapters.

Introduce an account-health subsystem with silent probes, projected
health/auth states, and IPC invalidation events so the renderer can
show accurate auth/probe status and verification timestamps.

Server-side, derive and forward title/business_date/scheduler_group_key/
question_text metadata on desktop task events so the local scheduler
can defer same-question fan-out before leasing.

task_plan.md

@@ -1,10 +1,10 @@
 # Task Plan: GEO admin-web backend completion and frontend foundation
 
 ## Goal
-Finish the completed backend verification/implementation chain, then resume the interrupted work by establishing the missing frontend foundation for `admin-web` so it can authenticate against `tenant-api` and render the first dashboard slice.
+Continue the desktop AI monitoring implementation by first adding a client-side local scheduler with durable queueing and adaptive execution policy, then adding a reusable hidden Playwright CDP execution layer that attaches to Electron Chromium and reuses existing account partitions.
 
 ## Current Phase
-Phase 14
+Phase 17
 
 ## Phases
 ### Phase 1: Progress Verification
@@ -94,10 +94,30 @@ Phase 14
 - [x] Deliver the extraction result and any remaining gaps
 - **Status:** complete
 
+### Phase 15: Desktop Scheduler Design & Integration
+- [x] Audit the current desktop runtime queue/task execution path and lock the scheduler constraints
+- [x] Implement a durable local monitor-task queue/cache with stale-task cleanup
+- [x] Implement platform-level mutual exclusion and adaptive global concurrency policy
+- [x] Integrate the scheduler into runtime leasing/execution without regressing publish tasks
+- **Status:** complete
+
+### Phase 16: Hidden Playwright CDP Execution Layer
+- [x] Add a reusable hidden-browser manager that attaches Playwright to Electron Chromium over CDP
+- [x] Reuse desktop account partition/session state for hidden pages
+- [x] Expose the hidden Playwright context/page lifecycle to future monitor adapters
+- [x] Keep the current hidden-view path working for existing adapters during the transition
+- **Status:** complete
+
+### Phase 17: Verification & Delivery
+- [x] Run targeted desktop type/build verification
+- [ ] Inspect runtime snapshot output for new scheduler/CDP state
+- [ ] Summarize what is production-ready versus still scaffolded
+- **Status:** in_progress
+
 ## Key Questions
-1. Which of the 9 steps are already implemented versus partially complete?
-2. What is the earliest step whose acceptance criteria are not yet met?
-3. What concrete code changes are required now to advance that step safely?
+1. How should the desktop client defer and locally optimize monitor-task execution without violating one-task-per-platform serialism?
+2. What is the smallest durable local cache that allows same-day resume while safely dropping stale next-day monitor tasks?
+3. How can Playwright CDP attach cleanly to Electron Chromium while reusing existing account session partitions?
 
 ## Decisions Made
 | Decision | Rationale |
@@ -115,6 +135,9 @@ Phase 14
 | Reconstruct the screenshot layouts from visual position data, not OCR text alone | The user explicitly asked to follow image design including element placement |
 | Move the quota card into the left sidebar footer and use page-level hero sections | This matches the reference screenshots more closely than the previous generic topbar layout |
 | Centralize prompt text in a dedicated package instead of leaving it inside business logic functions | The user explicitly asked to extract hard-coded prompts for easier future optimization |
+| Move scheduling authority fully into `desktop-client` | The user explicitly wants the server to dispatch tasks while the client decides when to execute them |
+| Allow stale cross-day monitor tasks to be dropped on the client | The user explicitly allows漏采 and does not want next-day catch-up for unfinished tasks |
+| Implement hidden browser infrastructure before expanding more adapters | Qwen and similar platforms cannot reliably use direct API calls and need browser-native execution |
 
 ## Errors Encountered
 | Error | Attempt | Resolution |
@@ -125,9 +148,12 @@ Phase 14
 | Gemini CLI `gemini-3.1-pro-preview` returned `429 MODEL_CAPACITY_EXHAUSTED` repeatedly | 1 | Stopped the Gemini path after the user requested not to use Gemini anymore |
 | Playwright MCP navigation failed locally because it attempted to create `/.playwright-mcp` | 1 | Fell back to preview HTTP checks instead of blocking the turn on Playwright environment setup |
 | Login button click did not trigger the sign-in flow in browser testing even though the backend endpoint worked | 1 | Bound the primary login button directly to `handleSubmit` and re-ran browser verification until the route change succeeded |
+| Desktop monitor execution is currently global-serial and lacks durable local queueing | 1 | Replace the ad-hoc in-memory FIFO with a scheduler module that owns persistence, concurrency, and stale-task cleanup |
+| `connectOverCDP()` is not wired yet even though `playwright-core` is installed | 1 | Add a dedicated hidden-browser manager instead of pushing CDP code directly into adapters or the runtime controller |
 
 ## Notes
 - Re-check task_plan.md before major implementation decisions.
 - Record concrete evidence for each claimed completed step.
 - Frontend scope for this turn is limited to the tenant-facing `admin-web` shell, not the platform ops console.
 - Visual references for this turn must be treated as layout guides, not just copy decks.
+- Desktop AI monitoring scope for this turn is infrastructure-first: scheduler plus hidden browser layer, not all six adapters.