Add UI critic and GUI developer agent workflow

docs/ui-iteration-loop.md

@@ -0,0 +1,57 @@
+# UI Iteration Loop (Lead + GUI Developer + UI Critic)
+
+This project uses a 3-role loop for steady UI improvement:
+
+1. Lead defines one iteration target.
+2. GUI Developer implements one focused improvement.
+3. UI Critic reviews and prioritizes issues.
+4. Lead decides: ship, revise, or run next iteration.
+
+## Lead Checklist Per Iteration
+
+1. Set a narrow goal (example: "Improve bookmark quick search usability on mobile").
+2. Define success criteria (2-4 points, testable).
+3. Assign implementation to `gui-developer`.
+4. Send produced diff to `ui-critic`.
+5. Resolve critic findings:
+   - blocking/high -> must fix before merge
+   - medium/low -> schedule for next iterations
+6. Start next cycle with exactly one new high-impact focus.
+
+## Handoff Templates
+
+Use these short templates to keep cycles fast.
+
+### Lead -> GUI Developer
+
+```
+Iteration goal:
+Scope:
+Acceptance criteria:
+Constraints:
+```
+
+### GUI Developer -> UI Critic
+
+```
+Goal:
+Files changed:
+Acceptance criteria status:
+Open questions:
+```
+
+### UI Critic -> Lead
+
+```
+Findings (blocking/high/medium/low):
+Residual risks:
+Recommended next focus:
+```
+
+## Guardrails
+
+- Keep diffs small and reversible.
+- Preserve existing app behavior outside the current scope.
+- Validate desktop and mobile behavior in every iteration.
+- Prefer measurable improvements over subjective redesign debates.
+