fix(subagent): keep fleet-dry parent mail to a digest plus blob

Author: TheGreatAxios

docs/ARCHITECTURE.md

@@ -148,7 +148,7 @@ A worker deny-on-ask registers a harness-owned denied-call envelope (`src/permis
 
 Two directors, selected by role:
 
-- **ChatDirector** (interactive, `src/agent/director.ts`) — Extends `DefaultDirector` with task list tracking, workflow nudges, LSP auto-activation, and multi-turn chat semantics. It never terminates the session: operator declines are surfaced as replies and the reactor stays alive for the next message. Yielding while a live fleet is running is allowed (idle-with-fleet); the open-task nudge does not rewrite that wait/reply, and the workflow idle rail does not declare the parent stuck while helpers run. The TUI idle-with-fleet seed alone is not occupancy: an empty fleet still auto-continues slash workflows. Occupancy, mailbox, and operator inbounds reset the workflow idle counter. When a worker finishes or fails while the parent is idle (including idle-with-fleet), occupancy delivers mailbox mail as system inbound so Skywalker starts a new turn without polling `wait_agents`. Mailbox mail is a Summary + Blockers digest (Findings when Summary is empty) with a blob pointer named for `read_file`, not the full report JSON. A failed blob write inlines a truncated report with an honest not-retrievable notice. When the fleet goes dry with tasks still todo/doing, the TUI runtime re-enters the parent with collected worker reports rather than settling idle. Already-collected IDs in that continuation are id/status/description only.
+- **ChatDirector** (interactive, `src/agent/director.ts`) — Extends `DefaultDirector` with task list tracking, workflow nudges, LSP auto-activation, and multi-turn chat semantics. It never terminates the session: operator declines are surfaced as replies and the reactor stays alive for the next message. Yielding while a live fleet is running is allowed (idle-with-fleet); the open-task nudge does not rewrite that wait/reply, and the workflow idle rail does not declare the parent stuck while helpers run. The TUI idle-with-fleet seed alone is not occupancy: an empty fleet still auto-continues slash workflows. Occupancy, mailbox, and operator inbounds reset the workflow idle counter. When a worker finishes or fails while the parent is idle (including idle-with-fleet), occupancy delivers mailbox mail as system inbound so Skywalker starts a new turn without polling `wait_agents`. Mailbox mail is a Summary + Blockers digest (Findings when Summary is empty) with a blob pointer named for `read_file`, not the full report JSON. A failed blob write inlines a truncated report with an honest not-retrievable notice. When the fleet goes dry with tasks still todo/doing, the TUI runtime re-enters the parent with the same digest for newly collected worker reports rather than settling idle. Already-collected IDs in that continuation are id/status/description only.
 
   Auto mode is toggled by CLI flags (`--auto` / `--no-auto`); there is currently no in-session key to toggle it (default on; constrained envelope — workspace writes and unconstrained shell auto-allow; installs, recursive rm, force/uncontained worktree changes, sensitive-path and opaque-wrapper shell still ask; contained non-force `git worktree add`/`remove`/`prune` and `list` auto-allow; shell file-mutation denied). It is not a separate edit/plan mode.
 

docs/IMPLEMENTATION.md

@@ -212,7 +212,7 @@ Listings are list-free, dumps are dump-locked: a bounded `ls`/`tree` prints name
 
 - **Alt+Enter** queues a follow-up (kind `"queue"`) delivered only on **session-idle** — parent-idle **and** no live fleet lanes (`run` goes idle). Session-idle Alt+Enter is a no-op. **Ctrl+C** stops the run.
 
-Idle-with-fleet is shipped: after a non-blocking `spawn_agent` dispatch the parent turn can settle while workers keep running. The runner emits a `fleet` event carrying the live-lane count (same liveness rule as the progress strip: interrupted leftovers are not live); the bridge holds the run busy on that count, so mid-hold Enter upgrades to a new primary turn (sent immediately) instead of queueing a steer, follow-ups keep waiting for true session-idle, and any steer left pending at the hold's engagement delivers immediately — the parent it was steering has already stopped. The workflow idle rail honors live helper occupancy, not the TUI idle-with-fleet seed: a live fleet is not a stuck parent, an empty fleet still auto-continues slash workflows, and occupancy / mailbox / operator inbounds reset the workflow idle counter. While the hold is up and the parent is not processing, occupancy flushes mailbox mail (`driveMailboxMail` + `buildMailboxMailMessage`) on store subscribe and idle-with-fleet settle — one child done while siblings run is enough. The mail payload is a Summary + Blockers digest (Findings when Summary is empty) with a blob pointer named for `read_file`, not the full report JSON; fleet-dry continuation lists already-collected IDs as id/status/description only and does not re-deliver an ID mailbox mail is already sending. Skip that shot when a fleet-dry open-task continuation is latched. Both wakes wait for send to settle as accepted before taking reports; a pending Promise is not a continuation, and a failed or uncertain send idles the parent. Exec-primary `wait_agents` mounts with no yield predicate (`createWaitAgentsTool({ sessions, fleetRecords })`) and blocks to ready/timeout/abort; parent interrupt/abort cancels that parked waiter via the tool AbortSignal (`timed_out`, workers stay running) rather than a second waiter table. `interrupt_agent` on a child unblocks wait as interrupted through the mailbox overlay. `shouldYieldWait` remains a supported-but-unwired tool option (no production mount passes it).
+Idle-with-fleet is shipped: after a non-blocking `spawn_agent` dispatch the parent turn can settle while workers keep running. The runner emits a `fleet` event carrying the live-lane count (same liveness rule as the progress strip: interrupted leftovers are not live); the bridge holds the run busy on that count, so mid-hold Enter upgrades to a new primary turn (sent immediately) instead of queueing a steer, follow-ups keep waiting for true session-idle, and any steer left pending at the hold's engagement delivers immediately — the parent it was steering has already stopped. The workflow idle rail honors live helper occupancy, not the TUI idle-with-fleet seed: a live fleet is not a stuck parent, an empty fleet still auto-continues slash workflows, and occupancy / mailbox / operator inbounds reset the workflow idle counter. While the hold is up and the parent is not processing, occupancy flushes mailbox mail (`driveMailboxMail` + `buildMailboxMailMessage`) on store subscribe and idle-with-fleet settle — one child done while siblings run is enough. The mail payload is a Summary + Blockers digest (Findings when Summary is empty) with a blob pointer named for `read_file`, not the full report JSON; fleet-dry continuation uses the same digest for newly collected terminals, lists already-collected IDs as id/status/description only and does not re-deliver an ID mailbox mail is already sending. Skip that shot when a fleet-dry open-task continuation is latched. Both wakes wait for send to settle as accepted before taking reports; a pending Promise is not a continuation, and a failed or uncertain send idles the parent. Exec-primary `wait_agents` mounts with no yield predicate (`createWaitAgentsTool({ sessions, fleetRecords })`) and blocks to ready/timeout/abort; parent interrupt/abort cancels that parked waiter via the tool AbortSignal (`timed_out`, workers stay running) rather than a second waiter table. `interrupt_agent` on a child unblocks wait as interrupted through the mailbox overlay. `shouldYieldWait` remains a supported-but-unwired tool option (no production mount passes it).
 
 `src/tui/stream-event-map.ts` maps reactor events onto the bridge's inbound events, and `src/tui/turn-state.ts` tracks the turn's status. `src/tui/turns-to-blocks.ts` hydrates a resumed session's stored turns into the same content blocks.
 

docs/TUI.md

@@ -725,8 +725,9 @@ follow-ups wait one more turn. Occupancy wakes consume mailbox reports only
 after send settles as accepted; a failed or uncertain send idles the parent
 so a later flush can retry. Mailbox mail is a Summary + Blockers digest
 (Findings when Summary is empty) with a blob pointer named for `read_file`.
-A failed blob write inlines a truncated report. Fleet-dry continuation lists
-already-collected IDs as id/status/description only. The mail and that fleet-dry continuation
+A failed blob write inlines a truncated report. Fleet-dry continuation uses
+the same digest for newly collected terminals and lists already-collected IDs
+as id/status/description only. The mail and that fleet-dry continuation
 are runtime-to-agent traffic — the fleet board owns worker status — so
 neither paints a transcript row, and neither rehydrates as one.
 

src/subagent/fleet-dry-drive.test.ts

@@ -529,13 +529,15 @@ describe("driveOpenTasksAfterFleetDry", () => {
       agent_id: string;
       status: string;
       description?: string;
+      summary?: string;
       report?: string;
     }[];
     expect(parsed).toEqual([
       {
         agent_id: "fresh",
         status: "done",
         description: "new lane",
+        summary: "fresh report",
         report: "fresh report",
       },
       {
@@ -620,13 +622,16 @@ describe("driveOpenTasksAfterFleetDry", () => {
       send: () => ACCEPTED_DELIVERY,
     });
     expect(driven).toBe(true);
-    const parsed = reportsJSONFromPrompt(sent[0] ?? "");
-    expect(Array.isArray(parsed)).toBe(true);
-    const report = (parsed as { report?: string }[])[0]?.report ?? "";
-    expect(report).toContain(
+    const parsed = reportsJSONFromPrompt(sent[0] ?? "") as {
+      report_uri?: string;
+      summary?: string;
+    }[];
+    expect(parsed[0]?.report_uri).toBe(
       `tool-output:///${fleetDrySpillKey("big", "report")}`,
     );
-    expect(report).not.toContain("TAIL-MARKER");
+    expect(sent[0]).toContain("use read_file with that URI");
+    expect(sent[0]).not.toContain("TAIL-MARKER");
+    expect(parsed[0]?.summary).not.toContain("TAIL-MARKER");
     expect(
       new TextDecoder().decode(
         store.blobs.get(fleetDrySpillKey("big", "report"))?.bytes ??

src/subagent/fleet-dry-drive.ts

@@ -15,9 +15,12 @@ import { parseSubAgentReport } from "./report.js";
 /** Enough of a lane report for a parent continuation; traces stay on disk. */
 export const FLEET_DRY_REPORT_CHARS = 8_192;
 
-/** Summary/Blockers inline in mailbox digest; the blob holds the rest. */
+/** Summary/Blockers inline in occupancy digest; the blob holds the rest. */
 export const MAILBOX_DIGEST_SECTION_CHARS = 2_048;
 
+/** Unstructured reports have no envelope; keep the inline teaser short. */
+export const UNSTRUCTURED_DIGEST_CHARS = 280;
+
 export const FLEET_DRY_CONTINUATION_PREFIX =
   "The fleet has gone dry. Remaining open tasks:";
 
@@ -206,9 +209,12 @@ export function fleetDrySpillKey(
   return `fleet-dry:${agentId}:${field}`;
 }
 
-function clipDigestSection(text: string): string {
-  if (text.length <= MAILBOX_DIGEST_SECTION_CHARS) return text;
-  return `${text.slice(0, MAILBOX_DIGEST_SECTION_CHARS - 1).trimEnd()}…`;
+function clipDigestSection(
+  text: string,
+  max = MAILBOX_DIGEST_SECTION_CHARS,
+): string {
+  if (text.length <= max) return text;
+  return `${text.slice(0, max - 1).trimEnd()}…`;
 }
 
 function mailboxSpillNotice(text: string, uri: string): string {
@@ -230,7 +236,9 @@ export async function spillWorkerField(
   field: "report" | "error",
   writeBlob?: FleetDryBlobWriter,
 ): Promise<string | undefined> {
-  if (text === undefined || writeBlob === undefined) return undefined;
+  if (text === undefined || text.length === 0 || writeBlob === undefined) {
+    return undefined;
+  }
   const key = fleetDrySpillKey(agentId, field);
   try {
     await writeBlob(key, new TextEncoder().encode(text), "text/plain");
@@ -260,6 +268,11 @@ export async function digestCollectedReport(
     report.report !== undefined
       ? parseSubAgentReport(report.report)
       : undefined;
+  const unstructured =
+    parsed !== undefined &&
+    parsed.findings.length === 0 &&
+    parsed.blockers.length === 0 &&
+    parsed.paths.length === 0;
   const reportUri = await spillWorkerField(
     report.report,
     report.agent_id,
@@ -274,18 +287,17 @@ export async function digestCollectedReport(
   );
   const summary =
     parsed !== undefined && parsed.summary.length > 0
-      ? clipDigestSection(parsed.summary)
+      ? clipDigestSection(
+          parsed.summary,
+          unstructured
+            ? UNSTRUCTURED_DIGEST_CHARS
+            : MAILBOX_DIGEST_SECTION_CHARS,
+        )
       : undefined;
   const findings =
     parsed !== undefined && summary === undefined && parsed.findings.length > 0
       ? clipDigestSection(parsed.findings)
       : undefined;
-  const blockers =
-    parsed !== undefined
-      ? clipDigestSection(
-          parsed.blockers.length > 0 ? parsed.blockers : "None.",
-        )
-      : undefined;
   const reportInline =
     report.report === undefined
       ? undefined
@@ -298,6 +310,14 @@ export async function digestCollectedReport(
       : errorUri !== undefined
         ? mailboxSpillNotice(report.error, errorUri)
         : clipDigestSection(report.error);
+  const blockers =
+    parsed === undefined || unstructured
+      ? parsed !== undefined && parsed.blockers.length > 0
+        ? clipDigestSection(parsed.blockers)
+        : undefined
+      : clipDigestSection(
+          parsed.blockers.length > 0 ? parsed.blockers : "None.",
+        );
   return {
     agent_id: report.agent_id,
     status: report.status,
@@ -494,13 +514,14 @@ export async function collectUncollectedTerminals(
   return reports;
 }
 
-export function buildFleetDryContinuationPrompt(
+export function buildFleetDryContinuationPrompt<T extends { agent_id: string }>(
   tasks: readonly Task[],
-  reports: readonly CollectedWorkerReport[],
+  reports: readonly T[],
 ): string {
   const open = tasks.filter(
     (task) => task.status === "todo" || task.status === "doing",
   );
+  const unique = dedupeByAgentId(reports);
   const taskLines = open
     .map((task) => `- ${task.id}: ${task.title} (${task.status})`)
     .join("\n");
@@ -509,7 +530,7 @@ export function buildFleetDryContinuationPrompt(
     taskLines,
     "",
     "Collected worker reports (already collected — do not call wait_agents for these agent_ids):",
-    JSON.stringify(dedupeByAgentId(reports)),
+    JSON.stringify(unique),
     "",
     "Continue the remaining work. Mark each task done or cancelled with manage_tasks",
     "when finished, or spawn_agent the next specialist. Do not end this turn while",
@@ -558,34 +579,37 @@ async function driveOpenTasksAfterFleetDrySpill(
       args.mailbox,
       args.lanes,
       false,
-      args.writeBlob,
+      undefined,
+      false,
     )
   ).filter((report) => !delivering.has(report.agent_id));
   const uncollectedIds = new Set(uncollected.map((report) => report.agent_id));
   const stubs = collectAlreadyCollectedStubs(args.mailbox, args.lanes).filter(
     (stub) =>
       !uncollectedIds.has(stub.agent_id) && !delivering.has(stub.agent_id),
   );
-  const reports = dedupeByAgentId([...uncollected, ...stubs]);
   const takeIds = uncollected.map((report) => report.agent_id);
   for (const id of takeIds) delivering.add(id);
-  const prompt = buildFleetDryContinuationPrompt(tasks, reports);
-  const takeReports = (): void => {
-    for (const id of takeIds) {
-      args.mailbox?.take(id);
-    }
-    releaseOccupancyDelivering(args.mailbox, takeIds);
-  };
   const fail = (): boolean => {
     releaseOccupancyDelivering(args.mailbox, takeIds);
     args.onSendFailure?.();
     return false;
   };
+  let prompt: string;
   try {
+    const digested = await digestCollectedReports(uncollected, args.writeBlob);
+    const reports = dedupeByAgentId([...digested, ...stubs]);
+    prompt = buildFleetDryContinuationPrompt(tasks, reports);
     args.beginSystemContinuation(prompt);
   } catch {
     return fail();
   }
+  const takeReports = (): void => {
+    for (const id of takeIds) {
+      args.mailbox?.take(id);
+    }
+    releaseOccupancyDelivering(args.mailbox, takeIds);
+  };
   return settleOccupancySend({
     send: () => args.send(prompt),
     onSuccess: takeReports,