feat(search): ⌃F searches the output across the panes

[HarryCordewener]

Third of a three-PR stack — based on feat/window-activity-boundary (#27), which is based on
feat/find-chords (#26). Review those first.
Design: docs/superpowers/specs/2026-08-11-pane-search-and-activity-design.md, part 2.

⌃F opens a modal results surface over the panes. Typing filters, ⌥E reads the query as a pattern,
⌥A widens from the focused window to every window, ⏎ goes to a hit and marks it with a bar, and
⌥G walks to the next one.

What is searched, and what is said about it

The pane buffer — not WorldSession.Scrollback, not the file-backed spill. RestoreLog's
reasoning one layer over: a spawn window's lines reach neither, so a session-keyed search would find
nothing in exactly the windows people search hardest. The bound is stated rather than implied
(12 found · 4,812 lines held), so a reader who cannot find an old line can see why.

Matching runs over the visible text, held per line and computed once at append. Over the markup, a
world changing colour mid-word would split a match — the same defect a wrapped URL has, one layer down
— and #ff0000 would find every red line. Chrome rows carry no plain text, so a search cannot find its
own bars.

Case is ignored in both modes (HistorySearch already is; (?-i) is the way back, which is why there
is no third toggle). An invalid pattern is a state, because a regex is unparseable most of the time
it is being typed, and there is a match timeout, because this runs on the UI thread per keystroke.

Landing

⏎ goes through Activate — the one activation path — so a hit in a background pane brings the pane,
the tab and the session forward together rather than scrolling something nobody is looking at. The bar
is the fourth of the boundary bars and points rather than repaints: the line is worth having because it
is the game's own text in the game's own colours.

Three things a reviewer might otherwise ask about

• No backward chord, and that was measured. ⌥⇧G decodes fine, but kitty writes it as
  CSI 103;4u — a kitty-keyboard-protocol sequence AnsiInputParser.DispatchCsi drops. ⌥⇧1's story
  one letter over. A decode test is not an arrival test.
• Escape does not clear the bar. A claimed Escape does not set _escapeAt, and TryAltEnter pairs
  an unclaimed one with a following Enter to make ⌥⏎. Binding it here would break the newline chord
  for as long as a bar was on screen.
• ⌥G removes the bar before re-running the search. The bar is itself a row, so a search run around
  it returns indices in a buffer about to lose one, and every hit below it lands a row early — which is
  what the test caught.

Two kinds of client chrome now live in the line buffers, so InsertChromeRow/RemoveChromeRow are the
one place that fixes up everything indexing into one. The demo scene also loads with _watching off:
it pours a spawn window's whole history in before the first frame, and every line otherwise counted as
missed.

Verification

dotnet build -c Release SharpMUTerm.slnx clean and warning-free; five suites green (2,722 tests).
Four new snapshot views rendered and looked at: search, search-regex, search-all,
search-landed (over a split, so the pane is narrower than the terminal — the geometry that catches a
landing scrolled to the wrong row).

🤖 Generated with Claude Code

https://claude.ai/code/session_015nuKnWthnELNkrd86q5KWN

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Pro

Run ID: 808e25f8-bd6c-4c9b-acca-582bff2fc1cb

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

_{Comment @coderabbitai help to get the list of available commands.}