## Thinking Path
> - Paperclip's UI is governed by the design system merged in #9134 and
the component convergence in #9240 — one Card, one Badge, one nav row,
one `IssueRow`, a single multiplicative radius ladder.
> - With those primitives in place, the remaining rough edges were
interaction and alignment details on the surfaces people use every day:
the inbox, the task list, the sidebar, and the routine/task detail
pages.
> - Each item here was reported from live use and fixed against a
running instance, then verified by measurement (pixel alignment, frame
timing) rather than by eye alone.
> - The result is that the inbox and task lists now behave as one system
(same hover, keyboard nav, tree-guide, and archive language), list
nesting reads correctly, list hover is smooth, and the routine detail
page scrolls and aligns like the rest of the app.
## Linked Issues or Issue Description
No public GitHub issue exists for this work; describing per the feature
template.
- **Problem**: after the design-system foundation (#9134) and component
convergence (#9240), the inbox/task lists still had interaction and
alignment gaps — hover lag on long lists, keyboard navigation that only
partly matched between the two lists, workspace/parent nesting whose
guides and chevrons didn't line up, an inbox that sat offset from the
task list, and a routine detail page with an odd double-scroll and a
bespoke sub-nav.
- **Proposed behavior**: the inbox and task lists share one interaction
contract (hover, keyboard nav, collapse, archive), list nesting aligns
to the status column with clean chevrons, list hover is CSS-only (no
per-hover re-render), and the routine detail page uses a fixed
header/sub-nav with a single scrolling content region and a sub-nav that
matches the primary nav.
## What Changed
- **List hover performance**: hover is now painted purely by CSS
`:hover` and records the hovered row in a ref, instead of writing
list-selection state on every `mouseenter` (which re-rendered 100–300
non-memoized rows per hover). Keyboard nav reads the ref so it still
continues from the hovered row; the keyboard band clears on the first
real mouse move so hover and keyboard selection never show two bands at
once. The row's `transition-colors` fade was removed so the highlight
snaps (no comet-tail). Measured on a 100-row scrub: worst frame **333 ms
→ 33 ms**, long frames (>50 ms) **12 → 0**, avg **41 → 60 fps**.
- **Inbox ↔ task-list parity**: keyboard navigation works on every inbox
tab (archive/read keys stay scoped to the archivable tab); group headers
and parent tasks collapse/expand with the arrow keys in both lists; the
task list gains the same j/k / arrows / Enter selection model as the
inbox; hover selection bands match; the `g` then `i` go-to-inbox chord
works app-wide.
- **List nesting & alignment**: the workspace group-header chevron lines
up exactly with the task chevrons below it (both lists); the
parent→child connector line drops from under the parent's **status
icon** rather than its chevron and breaks with a 14px gap around a
nested row's own chevron; and inbox rows line up with the task list
(read rows no longer reserve a mark-read column).
- **Inbox archive affordance**: moved from a bare `x` left of the status
icon to an `Archive` icon + label button on the right (before the
timestamp), revealed on row hover; the left slot now carries only the
unread dot; swipe-to-archive is unchanged.
- **Sidebar**: when no agent has a live run, the AGENTS section shows 3
recent agents (was 5) plus "See all agents"; the working-agents view is
unchanged.
- **Routine detail page**: the layout is bounded to the main scroll area
so the header (Run now + automation toggle) and the sub-nav stay fixed
and only the section content scrolls (was a page-level scroll competing
with a `sticky` sub-nav). The sub-nav items adopt the primary nav's
rhythm — row padding/height, inset rounded pill, type scale, and 16px
icons — and its background matches the main nav.
- **Task detail**: dropped the redundant `🔵` prefix the breadcrumb added
for live/in-progress tasks; the status glyph already conveys that state.
## Verification
- `pnpm check:token-gates` → 3/3 CLEAN
- `pnpm typecheck` → green (all packages)
- `cd ui && npx vitest run` → 2255/2255 (assertions updated in lockstep
where behavior changed)
- `pnpm --filter @paperclipai/ui build` → exit 0
- Storybook visual regression: run locally throughout (the
baseline-manifest archive is still unpublished, so CI cannot run this
suite — pre-existing condition from #9134). Every visible delta was
reviewed against a live instance and, where it was intentional (nesting
alignment, routine sub-nav restyle, unread-row shift), the affected
snapshots were re-baselined locally.
- Manual/measured: inbox + task lists (grouped and nested, light +
dark), hover-scrub frame timing, keyboard navigation, and the routine
detail scroll/alignment were exercised on a running instance.
## Risks
- Behavior-and-alignment changes concentrated in `IssueRow` /
`IssuesList` / `Inbox` (the shared task-row surfaces). The riskiest area
— the hover/keyboard-selection model — is covered by unit tests (updated
in lockstep) and was measured and driven live.
- The routine detail scroll change restructures that page's layout
container; verified the page's own scroll stays fixed while only the
section content scrolls.
- The visual suite cannot yet run in CI (unpublished baseline archive —
pre-existing); snapshot coverage is local-only until that lands.
## Model Used
- Claude Fable 5 (`claude-fable-5`, Anthropic) via Claude Code — agentic
coding with tool use (file editing, test execution, Playwright
measurement/screenshot verification); extended thinking enabled.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
53d09d4c34