a turn this tab did not start is still a turn in progress

"Turn finished · no closing message" kept appearing over an agent that was
plainly still working, with output resuming a moment later.

isGenerating was set in exactly three places: sendPrompt, and the two reconnect
handshakes. So it answered "am I waiting for a reply I asked for", never "is
this conversation working". Every other way a turn begins left it false — a
commander waking on its own schedule, a queued prompt the server delivered,
another tab driving the same session, or a turn that started before this tab
was looking. The events still render, but buildRows reads isGenerating to
decide whether the last turn is finished, so a live turn was `complete` and the
moment it sat on a completed tool call with nothing said after it, the banner
fired. Same gap left the composer enabled mid-turn, and a message typed then
was not marked injected — which splits the live turn and prints the banner a
second way, the exact failure injected exists to prevent.

session:claude now sets it: the harness's own turn-start signal, and the only
one that arrives before any work does. tool:start sets it too, because
session:claude is not durable and a socket connecting mid-turn never receives
one, while a tool cannot start outside a turn.

Two candidates were measured rather than argued. A genuinely silent turn ending
earns the banner honestly, but 667 of 673 recorded turns end with
assistant:text immediately before result, so that is ~1% and cannot be it. And
a stale sync:live is real but rare: correlating every attach in the officer log
against the agent's own later output, 15 of 549 "not generating" attaches were
contradicted and one provably so. That residue is left alone — the tool:start
fallback now corrects it on the next event — and noted on the refresh card,
which is the same seam.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Officer-Session: f7a62308-0977-4cf4-aa29-c020853fd420

TODO.md

@@ -21,10 +21,10 @@ the owner's OS user and can never be granted. Indirection there really is accide
 
 ## Themes
 
-**97 open, 30 closed**, across 32 themes.
+**97 open, 31 closed**, across 32 themes.
 
 | Theme | Open | Closed |
-| ------------------------------------------------------------------------------------------ | ---- | ------ |
+| --- | --- | --- |
 | [Affordances](todo/affordances/) | 2 | 0 |
 | [Agents coordinating agents](todo/agent-coordination/) | 3 | 0 |
 | [Speaking into a session from the backend](todo/agent-delivery/) | 1 | 1 |
@@ -37,7 +37,7 @@ the owner's OS user and can never be granted. Indirection there really is accide
 | [Fleet — one command across every server](todo/fleet/) | 5 | 2 |
 | [Headscale](todo/headscale/) | 0 | 5 |
 | [Infra (alpha)](todo/infra/) | 7 | 0 |
-| [Known bugs](todo/known-bugs/)                                                             | 5    | 3      |
+| [Known bugs](todo/known-bugs/) | 5 | 4 |
 | [Memory — one store instead of per-harness files](todo/memory/) | 1 | 0 |
 | [Mobile apps](todo/mobile-apps/) | 4 | 0 |
 | [Multi-user](todo/multi-user/) | 17 | 5 |

src/workspaces/officerdev/src/hooks/useChat.ts

@@ -409,10 +409,26 @@ export function useChat(initialSessionId?: string, initialModel?: string | null,
         break;
 
       case 'session:claude':
-        // The turn has only just begun and the conversation is already addressable. Everything that makes
-        // a refresh survivable hangs off this: the URL is what the reattach handshake sends, so writing it
-        // here rather than at `result` is the difference between "refresh mid-turn and lose the turn" and
-        // "refresh mid-turn and watch it carry on".
+        /*
+          A turn is starting, and THIS TAB MAY NOT BE THE ONE THAT STARTED IT.
+
+          `isGenerating` used to be set in exactly three places — `sendPrompt`, and the two reconnect
+          handshakes — so it answered "am I waiting for a reply I asked for", not "is this conversation
+          working". Every other way a turn begins left it false: a commander waking on its own schedule, a
+          queued prompt the server delivered, another tab or another device driving the same session.
+
+          The tab still rendered the work, because the events arrive either way. But `buildRows` reads
+          `isGenerating` to decide whether the last turn is finished, so with it false a live turn is
+          `complete`, and the moment it sat on a completed tool call with nothing said after it the reader
+          got "Turn finished · no closing message" — over an agent that carried on a second later. It also
+          left the composer enabled mid-turn, and a message typed then was not marked `injected`, which
+          splits the live turn: the precise failure `injected` exists to prevent (see types.ts).
+
+          Marked here rather than on the first tool call because this is the harness's own turn-start
+          signal, and it is the only one that arrives before any work does.
+        */
+        setIsGenerating(true);
+        setHasStarted(true);
         claudeSessionIdRef.current = msg.claudeSessionId;
         setClaudeSessionId(msg.claudeSessionId);
         if (replaceUrl) writePermalink(msg.claudeSessionId);
@@ -449,6 +465,9 @@ export function useChat(initialSessionId?: string, initialModel?: string | null,
       }
 
       case 'tool:start': {
+        // Belt and braces for a turn whose `session:claude` this tab missed — an event that is NOT durable,
+        // so a socket that connects mid-turn never receives one. A tool cannot start outside a turn.
+        setIsGenerating(true);
         toolCallsInTurnRef.current = true;
         const call: ChatMessage = {
           role: 'tool',

todo/known-bugs/a-turn-this-tab-did-not-start-reads-as-finished.md

@@ -0,0 +1,50 @@
+---
+id: known-bugs/a-turn-this-tab-did-not-start-reads-as-finished
+title: 'A turn this tab did not start reads as finished — "no closing message" over a working agent'
+theme: known-bugs
+status: done
+created: 2026-08-25
+date_source: stated
+order: 6
+---
+
+**"Turn finished · no closing message" appears while the agent is still working, and output resumes a
+moment later.** Reported by the owner 2026-08-25 with a screenshot: the session list shows Platform Harry
+`generating`, and Platform Harry's own panel shows the banner under two completed Bash calls. He reports it
+happens often.
+
+**Not the same bug as the two below it, and not a race.** `isGenerating` was set in exactly three places:
+`sendPrompt`, and the two reconnect handshakes (`sync:live`, `sync:messages`). So it answered _"am I
+waiting for a reply I asked for"_, never _"is this conversation working"_. Every other way a turn begins
+left it **false** — a commander waking on its own schedule, a queued prompt the server delivered, another
+tab or another device driving the same session, or simply a turn that started before this tab was looking.
+
+The tab still renders the work, because the events arrive regardless. But `buildRows` reads `isGenerating`
+to decide whether the last turn is finished (`generatingNow = isLast && isGenerating`), so with it false a
+live turn is `complete` — and the moment that turn sits on a completed tool call with nothing said after
+it, the banner fires. Then the next event arrives and the conversation carries on, which is exactly what
+"still outputting in a bit" is.
+
+Two further consequences of the same gap, both visible in the screenshot: the composer stays **enabled**
+during someone else's turn, and a message typed then is not marked `injected` — which splits the live turn
+and prints the banner a second way, the precise failure `injected` was added to prevent (`types.ts`).
+
+**Ruled out on the way, with evidence rather than reading.** A silent turn ending would earn the banner
+honestly, so that was measured first: across every durable event ever recorded, **667 of 673 turns end with
+`assistant:text` immediately before `result`**. Genuine silent endings are ~1% and cannot account for "a lot
+of times". A stale `sync:live` was the other candidate and is real but rare — correlating every attach in
+the officer log against the agent's own subsequent output, 15 of 549 "not generating" attaches were
+contradicted, and only **one** provably so (the first event after it was a `tool:result`, so a tool was
+already in flight when officer answered).
+
+---
+
+**Fixed 2026-08-25.** `session:claude` — the harness's own turn-start signal, and the only one that arrives
+before any work does — now sets `isGenerating`. `tool:start` sets it too, as a fallback: `session:claude` is
+not durable, so a socket that connects mid-turn never receives one, and a tool cannot start outside a turn.
+
+Residual, and left alone deliberately: `sync:live` can still report `isGenerating: false` for a turn that is
+running (the one proven case above, `sidecar.isClaudeGenerating` answering for an adopted session). The
+`tool:start` fallback now corrects that on the next event rather than leaving the turn wrongly settled, so
+the symptom is covered even where the answer is not. Fixing the answer itself belongs with
+[[known-bugs/refresh-mid-tool-call-renders-the-turn-folded]], which is the same seam.

todo/known-bugs/refresh-mid-tool-call-renders-the-turn-folded.md

@@ -43,3 +43,17 @@ Possibly related, same neighbourhood (refresh + a live turn + folding), not conf
 cause: `known-bugs/refreshing-a-finished-chat-turn-loses-output` — a different symptom (a turn that DID
 finish loses its output entirely and misreports `0 turns · 0 tok`) but the same refresh-time seam
 between what the transcript says and what the live socket confirms.
+
+**2026-08-25 — partly overtaken, and the measurement is worth having.** A separate cause was found and
+fixed for the same family of symptom: `isGenerating` was never set by anything except `sendPrompt` and the
+two handshakes, so a turn this tab did not start rendered as finished from the first event onwards. See
+[[known-bugs/a-turn-this-tab-did-not-start-reads-as-finished]]. `session:claude` and `tool:start` now set
+it, which means the last turn stops folding as soon as any work arrives — so the window this card is about
+is now bounded by the next event rather than open-ended.
+
+What is left here is the question this card actually asks, and it is now measured rather than suspected:
+correlating every attach in the officer log against the agent's own subsequent durable output, **15 of 549
+attaches that told a client "not generating" were contradicted, one of them provably** — the first event
+after it was a `tool:result`, so a tool was already in flight when `sidecar.isClaudeGenerating` answered.
+So `sync:live` does sometimes carry a wrong value, it is rare, and it is a wrong ANSWER rather than the
+timing gap this card assumed. `historicalCount` was not implicated.