pi-goal-list-loop-audit 0.34.4 → 0.34.6

package/extensions/loops/goal.ts

@@ -991,8 +991,21 @@ function heartbeatTick(): void {
     })
   ) {
     lastWedgeAlertAt = Date.now();
-    const msg = `${goalNoun()} appears wedged: no activity for ${Math.round((Date.now() - lastActivityAt) / 60_000)}m while the session is busy — likely a hung command (test/build/dev server without a timeout). Check the session; Esc kills a stuck tool call.`;
-    appendLedger(ctx.cwd, "wedge_alert", { silentMs: Date.now() - lastActivityAt });
+    // v0.34.5: a wedge while blocked on a subagent wait is a DIFFERENT animal
+    // from a hung bash command (junk-runner 2026-08-01: 2 Explore agents
+    // "thinking…" 31 minutes — working, but indistinguishable from hung
+    // without this hint). Name the wait and give the liveness check: a child
+    // whose tool-use counter stops moving is hung, not thinking.
+    const subWaits = new Set(
+      [...inFlightToolCalls.values()]
+        .filter((t) => t.name === "get_subagent_result" || t.name === "Agent")
+        .map((t) => t.name),
+    );
+    const subHint = subWaits.size > 0
+      ? ` The in-flight call is a SUBAGENT WAIT (${[...subWaits].join("/")}) — check the Agents panel: a child whose tool-use/token counters have stopped moving between checks is hung, not thinking (hard failures surface as ✗ failed + the wait returns; a HANG is silent). Esc interrupts the wait — then collect the survivors with get_subagent_result and absorb the dead scope inline.`
+      : "";
+    const msg = `${goalNoun()} appears wedged: no activity for ${Math.round((Date.now() - lastActivityAt) / 60_000)}m while the session is busy — likely a hung command (test/build/dev server without a timeout).${subHint} Check the session; Esc kills a stuck tool call.`;
+    appendLedger(ctx.cwd, "wedge_alert", { silentMs: Date.now() - lastActivityAt, subagentWait: subWaits.size > 0 });
     ctx.ui.notify(msg, "warning");
     notifyExternal(ctx, msg);
   }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-goal-list-loop-audit",
-  "version": "0.34.4",
+  "version": "0.34.6",
   "description": "Mission control for autonomous pi: interview-drafted goals, an audited task queue, and forever-loops (metric, spec, project-audit) that run for hours. An isolated extension-less auditor re-verifies every completion with raw evidence; confirmed drafts, decision pauses and consent gates keep you in charge.",
   "license": "MIT",
   "author": "dracon",

package/prompts/goal-loop-continuation.md

@@ -88,14 +88,16 @@ When the user says "do a full audit", "survey the project", "find all problems",
 4. Don't ship a single bug fix and then ask if the user wants to continue — the user already said "do a full audit".
 5. After the task list is confirmed, work through tasks systematically with `complete_task` / `update_task_status`.
 
-## WHEN SUBAGENTS DIE ON TOKEN LIMITS
+## WHEN SUBAGENTS DIE: RESUME, DON'T RESPAWN
 
-If a subagent fails with `run hit the output token limit` (or returns an error after many tool uses with no report): do NOT respawn the same wide brief — it will die the same way. Either:
+A failed subagent's SESSION survives its death (verified in pi-subagents source: `resume` has no status guard — it re-prompts the existing session, context intact). The 56 tool uses of work are NOT lost unless you throw them away. On any subagent failure — output token limit, quota, provider error:
 
-1. **Split it**: the brief was too wide — respawn as 2 narrower agents (half the subsystems each), or
-2. **Absorb it**: if the unscanned area is small, survey it inline yourself.
+1. **Resume it**: `Agent(resume: "<id>", prompt: "Wrap up now: report what you found within ~150 lines.")` — for quota deaths, wait for the quota window first (see below). A resumed agent already HAS its research; it only needs to report.
+2. **Resume failed with "not found"?** The manager forgot it (a /reload wipes the in-process registry — old IDs are unresumable). Only NOW respawn, and never the same wide brief: SPLIT into 2 narrower agents or ABSORB the remainder inline.
 
-A dead subagent's partial work is LOST — prevention (tight briefs, report caps) is cheaper than recovery.
+## AFTER A SESSION RESTART: YOUR SUBAGENTS ARE DEAD
+
+Subagents run IN-PROCESS — a /reload or restart kills every running one instantly and silently (no failure event, no ✗), and their IDs become unresumable. On a restored session, check your transcript for in-flight agents and relaunch them (tight briefs) or absorb their scope — do NOT sit waiting for results that can never arrive. This is why long fan-out passes belong under a glla goal/list item: the goal plane survives restarts and re-drives the fan-out through the continuation; an ad-hoc fan-out dies with the tab.
 
 ## WHEN SUBAGENTS HIT QUOTA ERRORS