Release 1.1.0

.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "downrange",
-  "version": "1.0.1",
+  "version": "1.1.0",
   "description": "A band, pane, and /downrange table: progress, ETA, and last output for Claude Code background shells and Monitor watches, tool activity for subagents, state for workflows, and the next /loop wakeup.",
   "author": { "name": "dot-agi" },
   "license": "MIT",

README.md

@@ -4,7 +4,7 @@
   <img src="https://img.shields.io/badge/Claude_Code-2.1.288%2B-D97757" alt="Claude Code 2.1.288+">
   <img src="https://img.shields.io/badge/Claude_Code-mod-555555" alt="Claude Code mod">
   <img src="https://img.shields.io/badge/license-MIT-78997A" alt="License: MIT">
-  <img src="https://img.shields.io/badge/version-1.0.1-555555" alt="Version: 1.0.1">
+  <img src="https://img.shields.io/badge/version-1.1.0-555555" alt="Version: 1.1.0">
 </p>
 
 <p align="center"><strong>Progress and an ETA for background shells and Monitor watches, read from their real output or past runs, with no model request.</strong></p>
@@ -144,7 +144,7 @@ Only dynamic `/loop` wakeups appear, and other schedules can clear or keep one.
 
 - **Only dynamic `/loop` wakeups.** Only wakeups that `ScheduleWakeup` creates appear. Fixed-interval loops such as `/loop 10m` and other cron schedules do not.
 - **Other schedules and the wakeup.** Downrange cannot tell which schedule fired a scheduled prompt, or which one-shot schedule in a turn-end snapshot is the wakeup. A prompt of a scheduled task, a routine, or a fixed-interval loop that fires from 5 seconds before the wakeup's due time on clears the wakeup, even before the wakeup's own prompt fires. A cancelled wakeup that only a snapshot can clear stays while any other one-shot schedule remains. It then shows until a scheduled prompt fires near its due time, a new `ScheduleWakeup` result replaces it, or it is 10 minutes past due.
-- **Progress read from output.** Progress is a guess from the text. A short date such as `10/11`, `3 of 7` inside a path or a version text such as `version 1.2 of 7.0`, or another ratio or percentage that is not progress, counts as progress and changes the bar and the rate ETA. Full dates such as `2026/10/11` and `10/11/2026` do not count.
+- **Progress read from output.** Progress is a guess from the text. A short date such as `10/11` counts as progress, unless a two-digit clock follows it, as in `10/11 14:03`. `N of M` followed by a space inside a path, such as `Disc 1 of 2 (live)/a.flac`, counts too, and so does another ratio or percentage that is not progress. These change the bar and the rate ETA. Full dates such as `2026/10/11` and `10/11/2026` do not count. Neither does `N of M` joined to a path or a number, such as `/backups/3 of 7/`, `version 1.2 of 7.0`, or `3 of 7.5 GB`. The date rule also drops real progress of 1 to 12 out of at most 31 before a clock, such as `3/15 00:12`.
 - **Only the end of an output.** A direct read of a file up to 4 MiB keeps the last 16,384 characters (UTF-16 code units), which can be up to 48 KiB of non-ASCII text. A tail read above 4 MiB keeps the last 16,384 bytes. Progress and the last line come from that end. Long lines after a progress line can push it out.
 - **ETAs during output silence.** When output stops, the bar keeps its last reading, and the rate ETA keeps counting down from the time of that reading. Once the estimate runs out, the ETA stays at `<1m` while the job runs. This holds after `quiet` shows too, so a row can show `20%`, `<1m`, and `quiet 2m 00s` together. A Monitor whose events stop behaves the same way.
 - **Monitor deadlines and rate ETAs.** A Monitor watch shows its deadline only when it has no progress-based ETA. Once its progress gives a rate ETA, or reaches 100% and shows `finishing`, the deadline no longer shows. A rate ETA can run past the deadline: a 30-minute watch that goes from 1% to 2% in its first minute shows `~1h 38m`, although its deadline is 29 minutes away.
@@ -155,12 +155,12 @@ Only dynamic `/loop` wakeups appear, and other schedules can clear or keep one.
 - **Large outputs on Windows.** Outputs over 4 MiB need `tail` on PATH. Without it, such a job shows `tail failed`. A launch or end note that the job already shows, such as `backgrounded by you`, stays in its place; `claude --debug` still shows `cannot tail the output of task <id>`.
 - **Tasks first seen in a snapshot.** Claude Code's turn-end snapshot can list a Monitor as a shell. A launch result is the result of the Bash or Monitor call that started a task. A Monitor first seen in a snapshot shows as a shell until its launch result, its first event, or its end notification arrives. Only its launch result gives it a deadline. A shell first seen in a snapshot shows no last line, progress, ETA, or `quiet` flag until its launch result or a notification names its output file. Only its launch result brings a usual-duration ETA. A task that started before Downrange loaded, such as one in a session moved to another process, gets no launch result. Claude Code's internal tasks also appear, under their engine labels, and a long-lived one can keep the band visible. Use **Hide** to dismiss the band.
 - **`awaiting confirmation`.** A turn-end snapshot that leaves out a job younger than 10 seconds does not end it. The job keeps running with `awaiting confirmation` until newer input confirms or ends it, such as a later snapshot or a notification. The label does not expire on a timer.
-- **Inputs that reach Downrange late.** Claude Code gives its events no shared sequence number. Downrange stamps each input with the time it receives it, and drops an input older than one it already applied to the same job. Hooks can overlap, so an end notification can land after a newer output read, agent list, or snapshot of its job. Downrange then drops the notification: the row keeps running until a later snapshot or agent list ends it. A shell row then shows `ended without a report` and no exit code. An input that Claude Code delivers late counts as new. For example, a Monitor event that arrives after you cleared its finished row brings the row back as running, with a `≥` age, until a later snapshot ends it.
+- **Inputs that reach Downrange late.** Claude Code gives its events no shared sequence number. Downrange stamps each input with the time it receives it, and drops an input older than one it already applied to the same job. Hooks can overlap, so an end notification can land after a newer output read, agent list, or snapshot of its job. A completed, failed, or stopped notification of a shell, Monitor, or workflow still applies then, with its exit code, because such a task never runs again. Downrange drops a late end of an agent or a teammate, a `blocked` end, and the end of a remote task such as a remote workflow, because such a task can run again under its id. That row keeps running until a later agent list ends it, or a snapshot ends it as `ended without a report`, with no exit code or end note. A teammate can leave the agent list as it ends, so its row ends at the next turn-end snapshot. A notification does not add back a job that left the list, such as one you cleared. It still can when it lands during `/clear`, `/resume`, or `/branch`, before Downrange restores what it remembers, or when Downrange forgot the task (see the 2,000 tasks below). The row then comes back as running, with a `≥` age, until a later snapshot ends it.
 - **Launch details that arrive late.** The same rule applies to launch results. A newer notification, Monitor event, snapshot, or output read of a task can land while Downrange still handles its launch result. Downrange then drops what the launch result adds. A Monitor then shows no deadline, and its age starts with `≥`. A shell can lose its usual-duration ETA, or its output file and launch note too, and that run adds nothing to the duration history. A notification that later names the shell's output file still brings back its last line.
 - **Commands that subagents start.** A command with the note `ends with its agent` ends when its subagent answers. It then shows `ended with its agent` and no exit code, unless its own notification arrives later.
-- **Agents that wait.** A subagent that the agent list reports as pending, waiting, or idle shows that word beside its tool count. It shows no `quiet` flag and no failed-calls flag, however long it stays in that state.
-- **Agents that run elsewhere.** Downrange sees the tool calls only of agents that run in this Claude Code process. A remote subagent keeps `0 tools` and shows `quiet` from 2 minutes after launch, even while it works. A teammate in a terminal pane of its own can show the same. Its notification or a snapshot still ends it.
-- **A teammate whose pane closed.** A teammate in a terminal pane of its own writes `running` or `idle` in its team's roster. When that pane closes or dies, the roster keeps the last word. Its notification or a snapshot ends its row, but a later agent list read shows it running again. That read comes at each turn end, and every second while another agent runs. The row can then switch between ended and running, keep the band visible, and show `idle` or `quiet` with no end note. Check the teammate's own pane before you act on the row. Use **Hide** to dismiss the band.
+- **Agents that wait.** A subagent that the agent list reports as pending, waiting, or idle shows that word, beside its tool count when it shows one. It shows no `quiet` flag and no failed-calls flag, however long it stays in that state.
+- **Agents that run elsewhere.** Downrange sees the tool calls only of agents that run in this Claude Code process. A remote subagent shows `remote`, and a teammate in a terminal pane of its own shows `own pane`. Their rows show no tool count and no `quiet` flag, so a stuck or dead one looks the same as a busy one. A row whose launch result Downrange missed, or that an earlier version of Downrange saved, still keeps `0 tools` and shows `quiet` from 2 minutes after launch, even while the agent works. Its notification or a snapshot still ends it.
+- **A teammate whose pane closed.** A teammate in a terminal pane of its own writes `running` or `idle` in its team's roster. When that pane closes or dies, the roster keeps the last word. Its notification or a snapshot ends its row, but a later agent list read shows it running again. That read comes at each turn end, and every second while another agent runs. The row can then switch between ended and running, keep the band visible, and show `idle` with no end note. Check the teammate's own pane before you act on the row. Use **Hide** to dismiss the band.
 - **A session moved to another process.** `/background` and a hand-off to the desktop app start a fresh process that resumes the conversation. Downrange starts empty there. Running tasks return at the next turn end, with `≥` ages and no last line. A shell's notification can still name its output file, end it, and show its last line. The `/loop` wakeup returns only with the next `ScheduleWakeup` result.
 - **Changes during `/clear`, `/resume`, or `/branch`.** Downrange carries the jobs and the wakeup as it last wrote them before the old session ended. An update that lands during the change, such as a launch or a notification, can be lost. A later turn-end snapshot then adds a lost running task, with a `≥` age, and ends a task that left its list, as `ended without a report`. Downrange remembers the last 2,000 tasks that left its list, and forgets the oldest first. A task does not count toward the 2,000 while it is back in the list. These commands carry that memory to the next session. A finished job leaves the list through **Clear finished**, through the 20-job limit of **Finished**, or when you leave its session with one of these commands, which carry only running jobs. After `/resume`, a job that the resumed session last showed running can return, if 2,000 other tasks left the list after it, counted across all later sessions. The job shows `ended without a report`, or shows as running until the next turn end.
 - **Another identity for the mod.** Claude Code keeps the duration history per plugin identity, `<plugin>@<marketplace>`. Installed from the `downrange` marketplace, Downrange is `downrange@downrange`. Loaded with `--plugin-dir`, it is `downrange@inline`. A copy under another identity, such as one installed from another marketplace, keeps a separate history, which starts empty. It shows no usual-duration ETA until it records runs of its own. The jobs and the wakeup live in session state under the plugin name `downrange`. A build under another plugin name cannot read them, so it starts with no jobs and no wakeup. A later turn-end snapshot brings running tasks back, with `≥` ages. The `/loop` wakeup returns only with the next `ScheduleWakeup` result.