agent: forward unhandled turn failures to manager

run_claude now keeps a 20-line stderr ring buffer and bails with it
inline (was just 'exit <status>'). agent serve loop, on Failed (not
PromptTooLong — that's already absorbed by drive_turn's compaction
retry), sends the error body to manager via the normal hyperhive
send. swallows transport errors — failure is already in journald
and the events sqlite. manager-only harness (hive-m1nd) is unchanged
so it doesn't try to notify itself.

hive-ag3nt/src/bin/hive-ag3nt.rs

@@ -166,6 +166,14 @@ async fn serve(
                 let outcome = turn::drive_turn(&prompt, files, &bus).await;
                 turn::emit_turn_end(&bus, &outcome);
                 bus.set_state(TurnState::Idle);
+                // Failures are unhandled by definition — PromptTooLong is
+                // absorbed inside drive_turn via compaction, so anything
+                // that reaches Failed here is a real crash. Notify the
+                // manager so it can investigate / restart / page the
+                // operator; best-effort, swallow the send error.
+                if let turn::TurnOutcome::Failed(e) = &outcome {
+                    notify_manager_of_failure(socket, e).await;
+                }
 
                 // After turn completes, check if there are pending messages waiting.
                 // If so, immediately process them instead of blocking on recv().
@@ -214,6 +222,29 @@ fn format_wake_prompt(from: &str, body: &str, unread: u64) -> String {
     format!("Incoming message from `{from}`:\n---\n{body}\n---{pending}")
 }
 
+/// Best-effort: tell the manager that this agent's last turn crashed
+/// (claude exited non-zero, compaction didn't help, etc.). Routed
+/// through the normal send path so the manager's inbox surfaces it
+/// like any other message; the agent's label is what the broker
+/// stamps as `from`, so the message body doesn't need to repeat it.
+/// Swallows transport errors — we just logged the failure, the worst
+/// case is the manager learns about the crash from the dashboard
+/// instead of inbox.
+async fn notify_manager_of_failure(socket: &Path, err: &anyhow::Error) {
+    let body = format!("claude turn failed:\n{err:#}");
+    let res = client::request::<_, AgentResponse>(
+        socket,
+        &AgentRequest::Send {
+            to: "manager".into(),
+            body,
+        },
+    )
+    .await;
+    if let Err(e) = res {
+        tracing::warn!(error = ?e, "failed to notify manager of turn failure");
+    }
+}
+
 /// Best-effort: ask our own per-agent socket how many messages are still
 /// pending after the wake-up Recv. Returns 0 if anything goes wrong.
 async fn inbox_unread(socket: &Path) -> u64 {

hive-ag3nt/src/turn.rs

@@ -3,6 +3,7 @@
 //! manager surface) and their wake-prompt wording; the spawn shape,
 //! arg-vector, stdin plumbing, and stream-json pumping are identical.
 
+use std::collections::VecDeque;
 use std::path::{Path, PathBuf};
 use std::process::Stdio;
 use std::sync::atomic::{AtomicBool, Ordering};
@@ -281,6 +282,13 @@ async fn run_claude(prompt: &str, files: &TurnFiles, bus: &Bus) -> Result<bool>
             }
         }
     });
+    // Keep the last STDERR_TAIL_LINES of stderr so a non-zero exit can
+    // include real context in the bail message (and downstream in the
+    // failure notification to the manager) instead of just "exit 1".
+    const STDERR_TAIL_LINES: usize = 20;
+    let stderr_tail: Arc<Mutex<VecDeque<String>>> =
+        Arc::new(Mutex::new(VecDeque::with_capacity(STDERR_TAIL_LINES)));
+    let tail_clone = stderr_tail.clone();
     let pump_stderr = tokio::spawn(async move {
         let mut reader = BufReader::new(stderr).lines();
         while let Ok(Some(line)) = reader.next_line().await {
@@ -293,6 +301,11 @@ async fn run_claude(prompt: &str, files: &TurnFiles, bus: &Bus) -> Result<bool>
             // surfaces when claude exits non-zero.
             tracing::warn!(line = %line, "claude stderr");
             bus_err.emit(LiveEvent::Note(format!("stderr: {line}")));
+            let mut t = tail_clone.lock().unwrap();
+            if t.len() >= STDERR_TAIL_LINES {
+                t.pop_front();
+            }
+            t.push_back(line);
         }
     });
 
@@ -301,7 +314,12 @@ async fn run_claude(prompt: &str, files: &TurnFiles, bus: &Bus) -> Result<bool>
     let _ = pump_stderr.await;
     let too_long = prompt_too_long.load(Ordering::Relaxed);
     if !status.success() && !too_long {
-        bail!("claude exited {status}");
+        let tail = stderr_tail.lock().unwrap();
+        if tail.is_empty() {
+            bail!("claude exited {status} (no stderr)");
+        }
+        let tail_str = tail.iter().cloned().collect::<Vec<_>>().join("\n");
+        bail!("claude exited {status}\nstderr tail:\n{tail_str}");
     }
     Ok(too_long)
 }