_smoothScrollBy and the touch paths call _maybeLoadMoreHistoryOnScroll on
every frame. Once the viewport sat at row 0, the next frame hit the real
pull's in-flight early return, so its reveal ran at once and the notice
showed the pre-pull tail state for the whole pull. A later frame now leaves
the reveal to the call that started the pull.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The response listened for 'data' and 'end' only. A socket closed after the
headers emits 'error' on the response and never 'end', so the promise never
settled. A one-shot `codeman agent` just exits, but the long-lived
`codeman mcp` left that tool call unanswered forever.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
After quick-start succeeded, a failed wait-output readiness call (a 429 from
the waiter cap, a network blip) became an isError result without the session
id, so the model could neither prompt nor delete the worker it had just
started, and a retry started a second one. It now returns the id with
ready: null and a note, as `codeman agent spawn` does.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
send_prompt and interrupt_session tagged input with seq = Date.now() under
one clientId. The server drops a seq at or below the last one it saw for that
clientId and still answers 200, so two inputs in one millisecond (parallel
tool calls) or a clock stepped back lost the second input while the tool
reported it sent. The seq is now max(last + 1, now).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Codex starts stdio MCP servers with a minimal environment, so `codeman mcp`
under the documented Codex config never learned the session it runs in: the
self-protection and the children rule did not apply. Both Codex TOML snippets
now forward the two variables with `env_vars`.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A shared-skills layout (`.agents/skills -> ../.claude/skills`) made
`codeman skill uninstall --case` report the symlink refusal as a failure and
exit 1 on every run, even with nothing left to remove. The refusal itself is
kept (nothing is written through the link); it is now reported as a skipped
directory.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
ExcelJS's worksheet pattern is unanchored, so several parsed worksheets can
share one sheetNo and therefore one rels list. repairExcelJsModel filtered and
indexed that list once per worksheet, which let a crafted admitted file run
past the 20 s preview timeout. Each distinct list is now filtered and indexed
once per sheetNo, and the relationship Type strings are built once.
SPREADSHEET_ASSET_VERSION is refreshed for the edited core.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Agents pick tools from MCP clients and registries, and Codeman had no MCP
surface: an assistant in Claude Desktop, Cursor, Codex or Gemini CLI could
not start or drive Codeman sessions without hand-written HTTP calls.
`codeman mcp` is a stdio MCP server and a CLIENT of the existing HTTP API
(it reuses codeman agent's transport and guards; no new route). Seven
tools: list_sessions, start_session, send_prompt, wait_for_session,
read_output, interrupt_session, delete_session.
- Dual-era: answers the legacy initialize handshake (2024-11-05 to
2025-11-25) and the 2026-07-28 revision (per-request _meta,
server/discover, resultType, and the ttlMs/cacheScope hints a modern
tools/list must carry: Claude Code 2.1.296 drops every tool without
them).
- Never guesses a server: the URL comes from --url or CODEMAN_API_URL.
Unconfigured, it still lists its tools and every call says how to
configure it.
- Prompting, interrupting and deleting are limited to sessions the process
started (or, inside a Codeman pane, that session's own workers) unless
--allow-any-session; never its own session. Control bytes are refused
before any request.
- Waits default to 45 s (Codex's MCP tool timeout is 60 s), max 600 s.
- read_output returns text only: a client that prefers structuredContent
otherwise showed the model metadata instead of the output.
- config/mcp-registry/server.json and package.json mcpName prepare the
official MCP registry listing; sync-plugin.mjs keeps its version.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The XLSX preview hands an admitted copy of the workbook to ExcelJS 4.4.0,
which links each worksheet to its comments, comment notes (VML), tables and
drawing by looking the relationship Target up verbatim in maps keyed by its
own spelling, then dereferences the result unchecked. openpyxl, the library
most agents write .xlsx files with, uses absolute Targets
(/xl/tables/table1.xml), its own comment part name (xl/comments/comment1.xml,
which ExcelJS never loads) and a drawing XML ExcelJS parses to nothing. One
comment, table or chart was enough for "Spreadsheet preview failed".
repairExcelJsModel() (spreadsheet-xlsx-core.js) runs on ExcelJS's parsed
model right before its reconcile: it re-points a link to the key ExcelJS
stored for the same part (so an openpyxl table is kept) and drops a link,
tablePart, drawing or hyperlink reference that would dereference nothing.
It adds nothing and resolves keys as own properties only. The worker wraps
workbook.xlsx.reconcile per instance to call it, and throws if a future
ExcelJS renames that method. SPREADSHEET_ASSET_VERSION is refreshed.
Tests: unit tests on model objects, and three real openpyxl 3.1.5
workbooks in test/fixtures/xlsx that fail the worker tests without the hook.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The agent skill only reached Claude Code: injection wrote
.claude/skills/codeman, which no other CLI reads, and the per-session
preamble was seeded for claude alone. A Codex, Gemini, OpenCode,
Antigravity, Grok, DeepSeek or OMP session inside Codeman never learned it
could start and drive workers.
- Registry field capabilities.agentSkillDir (a closed enum, since
clis.json can set it and it decides where the server writes) names the
skills directory each CLI scans. Claude keeps .claude/skills; those seven
CLIs get the same skill in <case>/.agents/skills/codeman (each measured to
scan it; Claude Code does not, so it never sees two copies). pi gets none:
a project skill makes pi ask "Trust project folder?" at startup.
- The preamble is seeded for every local session whose CLI takes the
skill; nothing in it is claude-specific.
- The skill text is CLI-neutral and names the two CLI-side failures to
report instead of retrying: a sandbox that blocks loopback (Codex's
default), and commands that do not carry this session's environment
(Codex 0.162 runs them in a shared app-server daemon).
- POST /api/sessions injects after the launch pre-flight, so a CLI that
cannot start leaves nothing in the workspace.
- codeman skill install gains --mode; uninstall --case removes the copy
from every directory the registry can write.
- The settings toggle moves to Agents & CLIs > All agent CLIs.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The agent skill section now shows a real run end to end: one short
prompt typed into Claude Code, the tile grid powering on with the CRT
entrance, then a DeepSeek Harness worker on a local qwen model and a
Claude Code worker powering on as new tiles, with lineage lines back to
the lead. 1280 wide, linking a 3600x2025 still.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The wiki said a session closed in the meantime frees its place for the
ranking. A session popped out into its own window does the same (it is
never tiled while its window owns its PTY size, and since the previous fix
popping it out with the grid open keeps the count, too), so the sentence now
names both.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The wiki, decision 11 in the tile-grid plan, its Persistence section, the
layout-memory test header and two tile-grid.js comments said the Tiles
button and a page reload bring the grid back however it was closed. Only
the Tiles button does: a grid closed before the reload stays stored as
open: false, the page shows the single view, and the toggle puts the grid
back. The behaviour was right; the wording now says that.
A test pins it through the real close path: arrange the grid, close it with
the toggle, start a fresh page, and the restore leaves the single view
while the toggle still brings back the cells, count, divider sizes, focus
and zoom. ('a stored closed grid leaves the single view' in
tile-grid-restore.test.ts covers the same from a raw stored value through
handleInit.)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A page reload puts a stored open grid back inside handleInit and fills a
cell freed since (its session deleted while the grid was closed) from the
ranking. Pending approvals, the source of the needs-input group, reach the
page only through seedApprovals, an async GET that lands after the restore,
so for that one fill a session waiting on a permission dialog or an unseen
finished turn ranks with the quiet ones (working sessions still rank first).
Holding the fill back until the seed lands would open fewer tiles, which
can be another shape, and then reshape the grid and move the user's tiles a
moment after the reload, so a late seed never re-forms a restored grid. The
plan, architecture-invariants#tile-grid and the wiki now say so, and a test
pins it: the seed arrives after the restore, the ranking then puts the
needs-input session first, and the restored cells and the stored layout
stay as they came back.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A session popped out while the grid was open (its pop-out button, or
another dashboard tab announcing the pop-out over the window channel) left
the grid through _markDetached, which removed its tile as if the user had
closed it: the stored count dropped to five and its cell stayed a hole the
ranking never filled, so after the window re-docked Tiles off and on showed
five tiles and an empty cell. Popping out with the grid closed, or a reload
reconcile, already kept the count and refilled the cell.
_markDetached now removes the tile as `gone` (it left by itself), the same
as a deleted session or a refused socket, so the count stays and the next
activation fills the cell from the ranking.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A page reload restores the stored grid inside handleInit; a cell freed by a
session gone since is filled from the ranking. handleInit syncs the tab
order and loads the session states before the restore runs, so the fill
ranks with real data: this test runs the real syncSessionOrder against a
server snapshot order and shows a working session taking the freed cell
ahead of the first tab. Pending approvals are seeded asynchronously after
init, so the needs-input group cannot rank on a reload.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The tile-grid plan records owner decision 11 (the layout is kept per
browser and comes back exactly; a fresh or filling grid ranks working,
then needing input, then most recent) and marks where it supersedes
decision 10's "the count wins over a remembered grid's size". The plan,
architecture-invariants#tile-grid and the CLAUDE.md Tile grid paragraph
now describe the stored format (`count`, freed cells, still v: 1), every
close path keeping it, and the open order; the wiki tells users what the
Tiles button brings back and how a new grid is chosen.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The Tiles toggle used to fill a remembered grid to the count picked in the
menu (refilling a hole the user made, or trimming tiles), removing the last
tile or "Open group as tiles" forgot it, and Ctrl/Cmd+click repacked it.
The owner asked for the opposite: "when I turn tiles off and on, always
keep what the last setting was".
Now the layout the user arranged (which session sits in which cell, holes
included, the tile count, the divider sizes, the focused tile and a zoom
they chose) is written to localStorage on every change and comes back
exactly from the toggle, Ctrl+Shift+G and a reload, however the grid was
closed (toggle, a non-tiled tab, leaveTiles or a #session= link, Home, the
width gate, the last tile, a group, closing or killing sessions). Nothing
of it reaches the server.
- The stored value stays v: 1 and gains `count`: how many tiles the user's
own last change left. A session that goes away by itself (deleted,
popped out, socket refused) does not lower it, so the next activation
fills that cell from the ranking; a hole the user made stays a hole.
- sanitizeTileGridState reports `freed` cells (sessions gone since) and
derives `count` for values written before it; restoreTileGridCells (pure)
puts a stored grid back, freed cells first, never trimmed to the window
(a too-small window shows the focused tile until it fits).
- Only when none of the stored sessions survive does activation rank from
scratch. A count picked in the menu still re-forms the grid to it.
- Nothing is persisted while a stored grid is being put back, and a
#session= link only flips `open`, keeping gone ids for the fill.
- The hover card no longer promises "a click opens N" when a stored grid
will open instead.
This supersedes decision 10's answer that the count wins over a
remembered grid's size; the affected tests say so.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The "Showing the most recent 1.0 MB of this session. 4.8 MB more may still
be retained." bar covered the top of the terminal on nearly every tab switch,
and for fullscreen claude panes it could never be acted on: `truncated`
measures the server's byte stream, while Load full history asks tmux, which
holds no scrollback for a pane that lives in the alternate screen.
- The capture reads `#{history_size}` in the cursor query it already makes
and the terminal route returns it as `paneHistoryLines`.
- A reported 0 hides the notice; a positive count replaces the byte gap in
the message.
- The notice is lazy: it appears once a wheel/touch gesture reaches the top
of the browser buffer (after the pull that gesture starts), leaves when
the user scrolls back to live output or switches tabs, and its dismissal
sticks per session until reload.
- formatHistoryBytes no longer prints "1024 KB" just under 1 MiB.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
npm takes the package's readme from an unsorted {README,README.*} glob at
publish time, and with README.zh-CN.md beside README.md it kept picking the
Chinese one, so npmjs.com rendered it. `npm run release` now runs
scripts/npm-release.mjs, which moves README.zh-CN.md to a dot-name for the
length of `changeset publish` and always puts it back (also on a failed or
thrown publish, and after an interrupted earlier run). It lives in the publish
command, not in release.yml, because changesets/action commits the working
tree into its version PR. Verified with npm's own normalize: README.zh-CN.md
before, README.md with the file aside, and the aside copy is never packed.
package.json: the description names the CLIs and says self-hosted (the old
"run 20 autonomous agents" was an internal performance target), and homepage
points at getcodeman.com instead of repeating the repository link.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The Tiles button with no stored grid opened the first tabs, so the oldest
sessions got tiles while the busy ones were left out. It now ranks the
open sessions: working first (the most recently started turn first, keyed
off lastSubmitAt only, since a working pane's activity stamp is always
"now"), then the ones needing input (the red and yellow tab alerts), then
everything else by most recent activity, tab order breaking ties.
The ranking is pure (rankTileSessions in constants.js) and reuses the home
screens' classification (_mobileOverviewState) and stamps. It feeds the
toggle's fresh set, a count picked in the menu, Ctrl/Cmd+click and an open
split's fill. The active session is still always included and focused,
detached sessions are never tiled, and a page without the classifier falls
back to tab order. The vm harness now loads mobile-overview.js so the app
tests exercise the real classifier.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Owner decision: searching for a tab and dragging it into a group is a useful
flow, so the drag-off guard is reverted and the docs and changeset say drag
works during a search.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
183efacc turned drag off while the rail search narrows the list, as the
owner asked, by refusing the flat manual rail's dragstart. That also
refused a drag that never reorders anything: with the tile grid open, a
found tab dragged onto a tile or an empty cell set no draggedTabId, so
_acceptTabDrops() (tile-grid.js) saw nothing and the drop was lost. On
the PR head that drop replaced or swapped the tile. The owner's reason
covers the reorder only: a reorder drop is saved for every device,
relative to rows the search hides, while the tile grid is per-device and
lands beside nothing hidden.
dragstart is unconditional again. The rows refuse the reorder instead:
their dragover returns before preventDefault while _tabRailSearchActive()
(the browser shows no-drop and fires no drop there), and their drop
returns before touching sessionOrder, for anything above the row that
might let a drop through. `draggable` stays on, so a cleared search
reorders with the same rows. The grouped rail's pointer drag keeps its
refusal in _onTabLayoutPointerDown(): it has no drop target besides the
rail.
Test: during a search that leaves both rows showing, dragstart is not
cancelled and sets draggedTabId, dragover on another row is not
accepted, a drop there leaves sessionOrder alone and saves nothing, the
real _acceptTabDrops() binder accepts the drag on a tile and hands it
the id, and after clearing the same rows reorder. It fails on the
previous commit, without the dragover guard, and without the drop guard.
CLAUDE.md, the invariants, the Dashboard wiki row and the release
changeset now say reordering by drag is off and a tile drop still works.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A floating subagent or ultracode window whose parent session's row the
rail search hid drew its connector from the viewport's top-left corner.
The hidden row is display:none, and getBoundingClientRect() still
answers it with an all-zero DOMRect, which is truthy, so the
`if (!tabRect) continue` paths never skipped it and _tabAnchor() put the
line at (0, 0) ("M 0 0 C 350 0, 350 400, 700 400" in Chromium). The
sidebar filter already had the same defect; #580 brought it to the rail,
and the keystroke redraw from 5f373a93 now ran it on every keystroke that
hides the parent.
A new _paintedSessionTab() in app.js returns the parent's row only while
getClientRects() is non-empty, and every place a floating window
measures its parent tab goes through it:
- the shared `tab:<id>` rect cache, filled by the subagent connectors
and by both ultracode connector layers (all three fill it, so guarding
only the first would let the next one cache the zero rect itself);
- the ultracode window's spawn position, which now cascades;
- the subagent window's fly-from-tab spawn, the same defect, not in the
report, now positioned as a window without a tab;
- the ultracode genie on minimize, which now tears down at once.
Lineage lines need nothing: they measure on a cache miss and computeTree
already drops zero-size rects.
Tests: in the gate, with the page laid out by hand (jsdom has no
layout), a subagent window, an ultracode run window and an ultracode
agent window from a hidden row draw no line and draw it again from the
row once the search is cleared, an ultracode window spawns from a
painted row and cascades from a hidden one, and the genie is skipped for
a hidden row. In the browser suite, real Chromium shows the hidden row's
rect is all zero and the connector is gone, then back after Clear. All
four fail on the previous commit. CLAUDE.md and the invariants'
Connectors bullet record the rule.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Closing a tab took ~0.55-0.7s on an idle machine, more with child
processes or recently active subagents. Most of it was waiting:
- The web UI kept the tab until DELETE returned, then removed it on the
100ms tab-render debounce, which every session update restarts.
closeSession() is now optimistic: the tab, tile and split go and the
next session is selected before the request is sent, rendered at
once. A refused delete (checked with a GET, since a delete can land
and lose its reply) puts the row back at its old index with the
error toast. This also fixes a latent bug: _apiDelete never throws,
so an HTTP error used to report "Session closed" while the session
kept running. SSE upserts skip ids that are being closed.
- The kill path slept fixed intervals (100ms PTY grace, 200ms for the
pane's children, 100ms for the process group) and verified in 100ms
steps. waitForProcessesExit() (utils/process-exit-wait.ts) keeps
every deadline but returns once the processes are gone, counting a
zombie as exited. Signal decisions keep kill(pid, 0).
- tmux kill-session and the pane-pid lookup ran via execSync, freezing
the server for ~70ms per close. Now async.
- killSubagentsForSession() ran a full `pgrep -f claude` scan per
active/idle subagent (~85ms each with ~100 matching processes). It
now scans once for all of them.
Measured on an isolated instance: click to tab gone 540-690ms -> 58-95ms;
DELETE of a claude session ~450ms -> ~200-260ms.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
CLAUDE.md gained a rule for the rail's Search sessions box and linked it
to architecture-invariants#session-list-layout-header-strip-vs-left-sidebar,
which said nothing about it, and the "Collapse is per-device" bullet under
Owner tab layouts had an exception it did not record.
- Session list layout: one paragraph on the shared _applyTabListFilter()
over the pure CodemanTabSearch (classes only, layout-scoped hide rules,
rail matches the name and the sidebar name + folder, locale-independent
lower-casing), the alert-row keep (owner decision), the data-total count
restore, the tree walk and roving-stop fix-up, the projection opening
every group, the connector redraw, the global Escape claim, no drag
while searching, and the reset off the rail.
- Owner tab layouts: the collapse bullet notes that a search draws every
group open and refuses toggles without writing the stored set.
- CLAUDE.md: the Escape claim and the connector redraw as one clause on
the existing rail search rule.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
From the owner's review of #580 rather than the bot's report: "Drag
while searching: I'd turn it off. A drop is saved for every device
(session order or the tab layout), and where it lands relative to the
rows the search is hiding is something you only see after clearing it."
The PR head left drag on.
Both rail drags now refuse while the search narrows the list: the
grouped rail's pointer drag in _onTabLayoutPointerDown(), and the flat
manual rail's HTML5 drag in its dragstart listener. The flat rail is
refused in the listener, not by flipping `draggable`, because a
keystroke in the box does not re-render the rows, so a cleared search
drags again with the same rows. The sidebar filter box, the header strip
and the keyboard moves (Ctrl+Shift+{ }, the row menu) are unchanged.
Tests: a grouped-rail press during a search starts no drag and one after
clearing does; a flat-rail dragstart during a search is refused and one
after clearing goes through. CLAUDE.md and the Dashboard wiki row say
so.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
CodemanTabSearch lower-cased the query and each row with
toLocaleLowerCase(), i.e. the browser's locale. Under a Turkish or Azeri
locale "API Review" lowers to "apı review", so a search for "api" missed
it, and the sidebar filter box, which used locale-independent
toLowerCase() before #580, changed matching with it. Both now use
toLowerCase(), like every other frontend search.
Test: a vm context of its own whose toLocaleLowerCase behaves like the
Turkish locale (the prototype is that context's alone) checks that "API"
still normalizes to "api" and that "API Review" matches. The test file's
overview also lists the pins the #580 fixes added.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Since the sidebar filter and the rail search share _applyTabListFilter(),
the sidebar's filter box also marks a .tab-cluster it emptied in the
by-case tab layout and rewrites its count. But the only rule hiding
.tab-cluster.tab-filtered-out was the rail-scoped one, so in the sidebar
the emptied box stayed painted with its label and a count of 0 while its
rows were hidden.
The sidebar rule now hides the marked case box as well. It stays scoped
to html[data-session-list="sidebar"], so a leaked class still cannot hide
anything on the header strip. An alerted row still counts toward its box
(the owner's call on #580), so a box holding a row that needs the user
never hides.
Tests: a unit test renders the sidebar in the by-case layout and checks
the emptied box is marked with a 0 count, that an alerted row keeps its
box on screen and counted, and that clearing restores the totals; a
stylesheet check pins the scoped selector; the browser test measures the
shipped styles.css in Chromium: the marked box is not painted in the
sidebar layout, an unmarked one is, and the header layout hides nothing.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A keystroke in the rail search box that does not change the group
structure only toggles classes, so hidden rows collapsed and the rows
below them moved up while the lineage lines and the subagent/ultracode
connectors kept pointing at the old positions until the next render. The
sidebar filter box had the same gap for its connectors.
_applyTabListFilter() now notes whether anything it touched actually
appeared or disappeared (a row, a group or case box, the state headings
via tabs-filtering, the "No sessions match" line) and calls
updateConnectionLines() only then. That covers both boxes and the alert
re-render, costs nothing on the re-apply every render tail runs when
nothing changed (the incremental path's lineage gate keeps its meaning),
and coalesces with a render's own redraw.
Same flag, owner's optional nit from the review: the grouped tree's
posinset/roving-stop fix-up at the filter tail now runs only while a
search hides something or right after one changed what shows, since both
render paths already set them over an unfiltered tree.
Tests: unit tests for the rail keystroke path (flat and grouped rail, no
render taken), the empty note, the sidebar box, and no redraw when
nothing moved; the browser test checks in Chromium that a class-only
keystroke moves a row up and redraws exactly once.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The global key handler in setupEventListeners() sits on document in the
capture phase, so it ran before the box's inline onkeydown: the Escape that
cleared the search also ran closeAllPanels() (collapsing the Monitor and
Subagents panels), closeHelp() and closeSessionManager(). Calling
stopPropagation() from the inline handler came too late.
The global Escape branch now claims an Escape whose target is
#tabRailSearch while the box holds text, the same way it already claims
one for the group menu, the grouped-rail drag and the Tiles count menu,
and routes it to handleTabRailSearchKeydown(). An empty box still leaves
Escape to the global handler, and an Escape that cancels an IME
composition is left to the IME.
Tests: the unit test installs the real global listener and asserts that
Escape with text in the box closes nothing, that an empty box closes as
before, and that an Escape outside the box is not claimed. The browser
test installs setupEventListeners() for real in Chromium, so the
capture-before-inline ordering is the shipped one.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Codeman maintainer