fix(session): persist an exit retraction, and let tests reach the watcher

Ten findings from a two-model review of this branch. Both reviewers cleared the
detection logic itself; everything here is a gap around it.

A route that starts a command in a pane now PERSISTS as well as broadcasts.
`/interactive` and `/shell` did neither before, and the pane-exit watcher cannot
cover for them: its next tick finds `paneExit` already cleared in memory,
reports no change and writes nothing, so `state.json` kept saying the agent had
exited for as long as the session stayed quiet. Nothing reads that record for a
decision yet, which is exactly why it had to be fixed now — part 2 is designed
to read it. The `clearPaneExitForNewPane()` docstring claimed its callers
already persisted; that claim was false for these two, and now says what the
caller owes instead.

The watcher's four guards were unreachable by any test. `refreshPaneExits()`
opened with `if (IS_TEST_MODE) return;`, so the read gate, the in-flight
suppression, the generation counter and the empty-read rule could each be
deleted with the whole suite green. The tmux call moves into `readPaneRows()`,
which a test subclass overrides — the shape `runRemoteReconnectTick` already
uses in this file for the same reason — and the test-mode gate moves with it, so
what a test cannot do is spawn a process rather than exercise the bookkeeping.
Each of the four guards now has a test that fails when it is deleted.

The muted status dot turned out to be a specificity fight on three surfaces, not
two. `.tab-status.error` was not excluded, so a session whose agent exited and
whose PTY-exit breaker then tripped lost its red dot to the mute — the state the
browser answers with a "restart it?" confirm, and a needs-you colour by the same
argument that protects the two alert classes. And mobile.css gives a `busy` dot
a 9px size and a green glow with `!important`, while `status` stays `busy` for a
pane whose agent died mid-turn, so a phone rendered a grey dot still wearing the
green halo beside a badge reading "exited". Both measured against the real
stylesheets, both now excluded, and the CSS test reads mobile.css too instead of
being structurally blind to half the problem.

Six comments said things that were not true. Two named the stats collector as
what replaces a restored reading, which is the opposite of the design. The
interval constant argued that 2000 ms keeps a read inside a tick, when the
5000 ms exec timeout means it cannot — which is why the in-flight guard exists.
`MuxSession.discovered` did not say the flag is permanent, though `saveSessions()`
serializes it. The empty-read docstring claimed a distinction that `|| true`
makes impossible. The invariants doc promised more than its drift test delivers.
And CLAUDE.md had no pointer at all, leaving its two hardest prohibitions
("never set `status: 'error'`", "never null the pid") only in the file it is
meant to route people to.

Refs Ark0N/Codeman#446.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

CLAUDE.md

@@ -203,6 +203,8 @@ Codeman is a Claude Code session manager with web interface and autonomous Ralph
 
 ⚠️ **A `❯` sighting is NOT the end of a turn, and neither is silence.** Claude redraws the composer (`❯`) about once a second all through a turn, so the old "saw a ❯, wait 2s → idle" rule flipped every working session to idle two seconds in (measured: a session mid-tool-call at 17 minutes reporting `status:"idle"`). Its working indicator is `✻ Actualizing… (13m 23s · ↓ 47.5k tokens)`: the glyph animates through `· ✢ ✳ ∗ ✻ ✽`, the gerund is randomized, and the finished line (`✻ Cooked for 2m 49s`) carries the same glyph, so neither `SPINNER_PATTERN` (braille, not what current versions draw) nor a keyword list can see it. Matching the new line in the STREAM does not work either: tmux ships partial repaints, so the whole line reaches the PTY only every few tens of seconds. So: `_confirmIdle()` (session.ts) requires the pane to go quiet, and then asks the SCREEN via `capturePaneText()` + `CLAUDE_WORKING_LINE_PATTERN` before believing it; a sustained run of repaints (`session-activity.ts`, pure + unit tested) is what marks a turn as started, with the same screen probe vetoing keystroke echo. Idle now lands ~3-5s after a turn ends instead of 2s into one. ⚠️ **The composer glyph and the working line are per-CLI registry DATA** (`capabilities.workDetect`, #385), not Claude constants: claude declares `❯` plus the pattern above, codex declares `›` plus `[Ee]sc to interrupt`, and a CLI that declares neither falls back to Claude's pair, which is what every session used before the registry carried one. Before that, this whole mechanism was gated Claude-mode-only on the reasoning that an external CLI has no `❯`, which was true and still left every Codex session reporting `idle` for its entire life. ⚠️ `workingLine` is config-supplied (a user `clis.json` can set it) and the compiled pattern runs on the PTY hot path, so it goes through `compileVersionRegex()` in BOTH the schema refine and `_workingLinePattern()`: a nested quantifier there is a ReDoS against the event loop, and the helper returns null rather than throwing so the fallback is structural.
 
+**An exited agent in a live pane** (`paneExit`, Ark0N/Codeman#446): Codeman creates every pane with `remain-on-exit on`, so `/exit` ends the CLI while tmux keeps the pane, the tmux session and the `tmux attach-session` process Codeman records as `Session.pid`. No PTY exit handler fires, so a session whose agent is gone reads as a live idle one. `TmuxManager.startPaneExitWatcher()` reads `#{pane_dead}`/`#{pane_dead_status}`/`#{pane_dead_signal}` from one batched `list-panes -a` on its OWN always-on interval, and `SessionState.paneExit` rides the existing `session:updated` broadcast. ⚠️ **Never set `status: 'error'` for an exited pane** — that value belongs to the PTY-exit breaker and makes the browser offer a restart — and **never null the `pid`**, which is what makes `selectSession()` re-attach and launch a fresh CLI. ⚠️ **The field is TRI-STATE and its third state is absence**, meaning UNKNOWN, which must never render as alive; `Session.paneExitApplies` is the one place that scoping lives and it fails closed for a direct-PTY session, a remote SSH session, a docker case and a record rebuilt from the socket. ⚠️ **An absent `#{pane_dead_status}` is not 0** (measured on tmux 3.2a, a SIGKILLed pane reports neither a status nor a signal), so never write `status ?? 0`: absent-stays-absent is what will keep a later clean-exit sweep off crashed agents. ⚠️ **A path that starts a command in a pane must clear the record AND persist**, since the watcher's next tick sees the field already cleared and writes nothing. → [architecture-invariants#an-exited-agent-in-a-live-pane-paneexit](docs/architecture-invariants.md#an-exited-agent-in-a-live-pane-paneexit)
+
 **Workspace-trust dialog auto-accept** (`session-trust-dialog.ts`, pure + unit tested): Claude Code asks once per directory ("Is this a project you created or one you trust?") before it will read or edit anything, and since Codeman sessions run permission-skipping or classifier-guarded modes the answer is always yes, so a session parked on that dialog is simply stuck. ⚠️ **Match the compacted SCREEN, never the stream.** tmux repaints a row by writing each word and then a cursor-forward (`\x1b[C`) instead of a space, and Ink colours each word separately, so the wire carries `I\x1b[Ctrust\x1b[Cthis\x1b[Cfolder`; stripping the escapes leaves `Itrustthisfolder`, because the spaces are not there to strip, they were never sent. A plain `includes('trust this folder')` therefore never matched a single chunk and the auto-accept was silently DEAD for every session that hit the dialog. `compactScreenText()` removes ALL whitespace instead (plus the `ESC ( B` charset selects that `stripAnsi` does not cover, which would otherwise land inside a phrase as a literal `(B`), which survives both that repaint style and the spaced full-screen redraw. ⚠️ **Never answer it with a blind `\r`.** The layout has changed under us at least twice, and Claude Code 2.1.252 dropped the option numbers, put "No, exit" FIRST and highlights IT by default, so the Enter that answered the old dialog now picks *exit* and the pane dies (`Pane is dead (status 1)`) seconds after the session starts. `trustDialogNextKey()` reads the `❯` marker and returns ONE step at a time (an arrow while the cursor is on the wrong option, Enter only once the screen shows it on the trust option), with the pane re-read between steps, so a dropped arrow costs a repaint instead of the session; a frame that does not say which option is highlighted returns null and waits for the next repaint. ⚠️ The LAST marked option in the text wins, because the direct-PTY fallback reads an append-only buffer where every repaint since launch is still present and an older frame must not out-vote the freshest one. ⚠️ Answering types into a live session, so THREE guards must all hold and none is redundant: a **startup-only window** (`TRUST_DIALOG_WINDOW_MS`, 90s, since the dialog renders before the main UI and leaving it open forever would let an agent transcript that merely QUOTES the dialog trigger an Enter, this file being an example), a **two-marker match** requiring a trust phrase AND one of the dialog's own confirm affordances (`isTrustDialogScreen`), and an **attempt cap** (`TRUST_DIALOG_MAX_ATTEMPTS`, 6: a keystroke can land while Ink is still mounting the widget and be dropped, which is the other half of why sessions got stuck here, but retrying forever would hammer keys into whatever came next; it was 3 while one Enter answered the dialog, and answering now costs at least two keystrokes). ⚠️ It reads `capturePaneText()` and falls back to a deliberately SHORT tail of the terminal buffer only on a direct-PTY session, which has no pane: that buffer is append-only, so a longer tail would keep re-matching a dialog answered minutes ago. ⚠️ **The scan must schedule its own next read** (`_trustDialogTimer`, cleared in `_clearAllTimers()`): it runs from the PTY `onData` handler, which was enough while one Enter answered the dialog, but the arrow that moves the cursor is the LAST output the pane produces, so a two-keystroke answer waiting on more output parks forever with the cursor sitting on the right option (measured on a live 2.1.252 spawn: cursor moved at 6 s, then nothing).
 
 **Process-tree walks are bounded** (`proc-tree.ts`, pure + unit tested): `collectDescendants(pid, byParent)` is the ONE descendant traversal, fed by a single cached `ps -eo pid=,ppid=` snapshot (`refreshProcSnapshot()` in tmux-manager.ts: in-flight-shared, async because `execSync`'s timeout cannot return at all while spawnSync waits on an unkillable child, and ANY error discards the result rather than caching a truncated `ps`, which would make whole subtrees invisible to the kill path). ⚠️ **The unbounded version took a machine down** (2026-07-30): it ran `pgrep -P <pid>` once per node and recursed with no visited set, no depth limit and no node cap, so across ~28 adopted tmux trees the fan-out exploded while each `pgrep` blocked in the WSL kernel reading `/proc/<pid>/cgroup`, ending at ~13,000 `pgrep` processes in D-state, a load average above 13,000, and a machine recoverable only by restarting WSL, which cost every running session. Three properties make that impossible and each has a test: a cycle terminates (a real tree has none, a stale snapshot can still produce one), depth is capped (`PROC_WALK_MAX_DEPTH`), node count is capped (`PROC_WALK_MAX_NODES`). The fourth is structural: the function takes a snapshot and cannot spawn anything at all. ⚠️ It lives in its own module because as a private method of `tmux-manager.ts` the regression test had to keep its own COPY of the algorithm, which is a test that passes while the shipped code rots. ⚠️ Truncation is reported through `onTruncated` rather than silently, with BOTH caps named: a silent depth cap hides a deep tree exactly as effectively as a silent node cap hides a wide one.