- Skip re-append for cards already in stack to avoid replaying enter animation
- Track enter animation end to remove `entering` class after first render
- Gate mouseenter interactive state on cards with explicit actions only
- Bind interactive hover only to action buttons and their close buttons
- Add regression tests for passive vs action-bearing notification hover
- New toggleNotificationHistory (Ctrl+N) session-scoped history panel; slides in from same edge as notification stack
- Overlay error/recovery toast follows notifications.overlayPosition; stack and history side seeded at startup
- Cold managed background startup initializes tray and visible overlay shell before tokenization warmups finish
- Add Update button to overlay update-available notifications
- Fix Ctrl+S sentence-card flow: only Anki progress notification, no duplicate status toast
- Fix overlay notification close/actions clickability above subtitle bars on Linux
- Increase pause-until-ready default timeout from 15s to 30s
Cross-check every config key, shortcut, default, and command against the
current source and fix the drift (mpv.socketPath, auto_start_overlay
default, AniSkip TAB key, JLPT N4 color, secondary-sub font/defaults,
secondary-sub language behavior, modular mpv plugin layout, and more).
Add plain-language intros and first-use definitions across onboarding and
integration pages so non-technical readers can follow along.
Internal docs/: fix stale module paths in architecture/domains.md, add
missing contract entry points and catalog rows, and bump verified dates.
Remove the obsolete docs/plans/ directory (its only plan shipped in
0.15.0) and reframe planning.md so plans live with the work, not in docs/.
- Replace `renderGroupedChanges` with `polishFragmentsWithClaude` that pipes fragments through `claude -p --model sonnet` to merge related items, drop housekeeping noise, and produce user-facing release notes
- Internal fragments kept in CHANGELOG.md under a `<details>` collapse; dropped from GitHub release notes entirely
- CI no longer auto-runs `changelog:build` on tag-based releases — fails fast with a clear error if `changes/*.md` fragments are still pending; build locally and commit before tagging
- Add `runClaude` dep-injection seam to test surface; add failure-mode coverage (missing binary, empty output, missing headers, missing `<details>` wrapper)
- Delete implemented design doc; update `changes/README.md` and `docs/RELEASING.md` with claude CLI prerequisite and new workflow
- Capture decisions from brainstorming: replace bullet renderer with `claude -p`, write straight to disk, hard-fail on missing/failed claude, drop internal section from release notes but keep collapsed in CHANGELOG.md
- Document prompt input/output contract, affected files, test plan, and CI guard that fails tag-based releases when changelog fragments are still pending
- Set scope boundaries (no caching, no SDK fallback, no `--no-polish` escape hatch)
- Add Arch Linux `subminer-bin` install instructions to README and docs-site installation guide
- Add Hyprland troubleshooting guidance for transparency and global shortcut passthrough
- Add TASK-159..164 backlog records and skill design plans; include changelog fragment
- Apply formatting-only update to `src/release-workflow.test.ts`
sudacode
