Propose a clean target model for view responsibilities:
- one-line purpose per view/tab,
- key interactions per surface,
- interactions that must be shared vs isolated.
### 3) Prioritized execution backlog
Create beads for follow-up work from critique findings:
- one bead per coherent unit,
- include scope/out-of-scope and acceptance criteria,
- preserve dependency correctness.
### 4) Minimal change strategy
Recommend a staged rollout plan that avoids large regressions:
- phase 1: low-risk high-value consistency fixes,
- phase 2: IA/view role cleanup,
- phase 3: deeper workflow refinements.
## Constraints
- Preserve current route model (`/` with `view=` query params).
- Keep changes grounded in actual implemented code (no speculative claims).
- Reuse shared components/logic; avoid one-off behavior per view.
- Keep language simple and operator-facing.
- **Approval gate:** Do not create any beads during discovery/brainstorming. First present findings + proposed bead backlog draft, then wait for explicit user approval before running any `bd create` commands.
## Quality bar
The critique should read like a professional product/UX architecture review for an agent operations platform, not generic UI feedback.
## Completion criteria
- Clear diagnosis of current UX shape using actual uncommitted code.
- Decision-ready target model for views/tabs and communication surfaces.
- Prioritized, execution-ready bead backlog generated from findings.
