@integrity-labs/agt-cli 0.28.450 → 0.28.452

package/dist/bin/agt.js

@@ -40,7 +40,7 @@ import {
   success,
   table,
   warn
-} from "../chunk-NJZI6BPI.js";
+} from "../chunk-WOZINHUT.js";
 import {
   AnchorSessionClient,
   CHANNEL_REGISTRY,
@@ -4830,7 +4830,7 @@ import { execFileSync, execSync } from "child_process";
 import { existsSync as existsSync10, realpathSync as realpathSync2 } from "fs";
 import chalk18 from "chalk";
 import ora16 from "ora";
-var cliVersion = true ? "0.28.450" : "dev";
+var cliVersion = true ? "0.28.452" : "dev";
 async function fetchLatestVersion() {
   const host2 = getHost();
   if (!host2) return null;
@@ -6002,7 +6002,7 @@ function handleError(err) {
 }
 
 // src/bin/agt.ts
-var cliVersion2 = true ? "0.28.450" : "dev";
+var cliVersion2 = true ? "0.28.452" : "dev";
 var program = new Command();
 program.name("agt").description("Augmented CLI \u2014 agent provisioning and management").version(cliVersion2).option("--json", "Emit machine-readable JSON output (suppress spinners and colors)").option("--skip-update-check", "Skip the automatic update check on startup");
 program.hook("preAction", async (thisCommand, actionCommand) => {

package/dist/{chunk-NJZI6BPI.js → chunk-WOZINHUT.js}

@@ -4910,7 +4910,7 @@ function requireHost() {
 }
 
 // src/lib/api-client.ts
-var agtCliVersion = true ? "0.28.450" : "dev";
+var agtCliVersion = true ? "0.28.452" : "dev";
 var lastConfigHash = null;
 function setConfigHash(hash) {
   lastConfigHash = hash && hash.length > 0 ? hash : null;
@@ -7411,4 +7411,4 @@ export {
   managerInstallSystemUnitCommand,
   managerUninstallSystemUnitCommand
 };
-//# sourceMappingURL=chunk-NJZI6BPI.js.map
+//# sourceMappingURL=chunk-WOZINHUT.js.map

package/dist/lib/manager-worker.js

@@ -42,7 +42,7 @@ import {
   resolveEffectivePinRaw,
   safeWriteJsonAtomic,
   setConfigHash
-} from "../chunk-NJZI6BPI.js";
+} from "../chunk-WOZINHUT.js";
 import {
   getProjectDir as getProjectDir2,
   getReadyTasks,
@@ -841,6 +841,14 @@ function logClaudeMdCeilingTransition(codeName, report, log2) {
 function reaperRestartBreakerReason(activeKeys) {
   return activeKeys.length >= 2 ? "mcp-presence-reaper" : void 0;
 }
+var KNOWN_REASON_CLASS_MAP = {
+  crash: true,
+  provisioning: true,
+  "credential-rotation": true
+};
+var KNOWN_REASON_CLASSES = new Set(
+  Object.keys(KNOWN_REASON_CLASS_MAP)
+);
 var PROVISIONING_RELOAD_REASONS = /* @__PURE__ */ new Set([
   "hot-reload-mcp",
   "managed-mcp-churn",
@@ -897,6 +905,11 @@ function maxProvisioningBucket(events) {
   }
   return best;
 }
+function tripClass(trip) {
+  if (trip.trippedClass) return trip.trippedClass;
+  if (!Array.isArray(trip.eventsAtTrip) || trip.eventsAtTrip.length === 0) return "crash";
+  return trip.eventsAtTrip.every((e) => isProvisioningReloadReason(e.reason)) ? "provisioning" : "crash";
+}
 var DEFAULT_MAX = 2;
 var DEFAULT_WINDOW_MS = 6e5;
 var DEFAULT_PROVISIONING_MAX = 5;
@@ -1007,7 +1020,10 @@ var RestartBreaker = class {
       const trip = {
         trippedAt: at,
         eventsAtTrip: [...tripEvents],
-        statusMessage: formatStatusMessage(tripEvents, tripWindowMs, trippedClass, provBucket.key)
+        statusMessage: formatStatusMessage(tripEvents, tripWindowMs, trippedClass, provBucket.key),
+        // ENG-7577: persist the tripping class so the recovery path can gate on
+        // it after a manager restart, without re-deriving it from prose.
+        trippedClass
       };
       this.trips.set(codeName, trip);
       this.events.delete(codeName);
@@ -1048,7 +1064,11 @@ var RestartBreaker = class {
     if (!saved) return;
     for (const [codeName, trip] of Object.entries(saved)) {
       if (trip && typeof trip.trippedAt === "number" && Array.isArray(trip.eventsAtTrip)) {
-        this.trips.set(codeName, trip);
+        const klass = trip.trippedClass;
+        this.trips.set(
+          codeName,
+          klass === void 0 || KNOWN_REASON_CLASSES.has(klass) ? trip : { ...trip, trippedClass: void 0 }
+        );
       }
     }
   }
@@ -1275,6 +1295,9 @@ function decideAutoResume(opts) {
   const { trip, marker, config: config2, now } = opts;
   if (!config2.enabled) return { action: "skip", reason: "disabled" };
   if (!trip) return { action: "skip", reason: "no-trip" };
+  if (tripClass(trip) !== "provisioning") {
+    return { action: "skip", reason: "class-not-eligible" };
+  }
   if (marker && now - marker.autoResumedAt < config2.backoffResetMs) {
     return { action: "skip", reason: "already-auto-resumed" };
   }
@@ -1296,6 +1319,28 @@ function hydrateAutoResumeMarkers(saved) {
   return map;
 }
 
+// src/lib/breaker-stale-trip.ts
+var DEFAULT_STALE_TRIP_SETTLE_MS = 12e4;
+var MIN_STALE_TRIP_SETTLE_MS = 1e3;
+function readStaleTripSettleMs() {
+  return Math.max(
+    MIN_STALE_TRIP_SETTLE_MS,
+    readEnvNumber("AGT_RESTART_BREAKER_STALE_TRIP_SETTLE_MS", DEFAULT_STALE_TRIP_SETTLE_MS)
+  );
+}
+function decideStaleTripClear(opts) {
+  const { trip, agentStatus, killSwitched, settleMs, now } = opts;
+  if (!trip) return { action: "hold", reason: "no-trip" };
+  if (agentStatus !== "active" || killSwitched) {
+    return { action: "hold", reason: "pause-in-force" };
+  }
+  const ageMs = now - trip.trippedAt;
+  if (ageMs < settleMs) {
+    return { action: "hold", reason: "settling", remainingMs: settleMs - ageMs };
+  }
+  return { action: "clear", ageMs };
+}
+
 // src/lib/resume-reconciler.ts
 var DEFAULT_MIN_HEALTHY_CYCLES = 2;
 var DEFAULT_BACKOFF_RESET_MS2 = 864e5;
@@ -1726,12 +1771,22 @@ var WINDOW_EXEMPT_RESTART_REASONS = /* @__PURE__ */ new Set([
   "sender-policy-change",
   "channel-mcp-down"
 ]);
+var RESTART_WINDOW_DEFER_ESCALATE_AFTER = 20;
+function shouldRelaxRestartWindowGate(deferStreak) {
+  return deferStreak >= RESTART_WINDOW_DEFER_ESCALATE_AFTER;
+}
+function nextWindowDeferStreak(priorStreak, decision) {
+  if (decision === "proceed") return 0;
+  if (decision === "defer-window") return priorStreak + 1;
+  return priorStreak;
+}
 function isWindowExemptRestartReason(reason) {
   return reason != null && WINDOW_EXEMPT_RESTART_REASONS.has(reason);
 }
 function decideRestartGate(opts) {
   if (isMaintenanceWindowDisabled(opts.env)) return "proceed";
-  if (!opts.windowExempt && opts.window && !isWithinMaintenanceWindow(opts.window, opts.now)) {
+  const windowAxisApplies = !opts.windowExempt && !shouldRelaxRestartWindowGate(opts.windowDeferStreak ?? 0);
+  if (windowAxisApplies && opts.window && !isWithinMaintenanceWindow(opts.window, opts.now)) {
     return "defer-window";
   }
   const paneThreshold = opts.idleThresholdSeconds ?? RESTART_IDLE_THRESHOLD_SECONDS;
@@ -8895,6 +8950,33 @@ function maybeReportCurrentTrip(codeName) {
     log(`[circuit-breaker] Failed to report trip for '${codeName}' to API: ${err.message}`);
   });
 }
+function maybeClearStaleTrip(agent) {
+  const codeName = agent.code_name;
+  const trip = restartBreaker.getTrip(codeName);
+  const decision = decideStaleTripClear({
+    trip,
+    agentStatus: agent.status,
+    killSwitched: Boolean(agent.kill_switch),
+    settleMs: readStaleTripSettleMs(),
+    now: Date.now()
+  });
+  if (decision.action !== "clear") return;
+  restartBreaker.clear(codeName);
+  reportedTrips.delete(codeName);
+  dependencyRecoveryLedger.clear(codeName);
+  log(
+    `[circuit-breaker] agent=${codeName} cleared STALE trip \u2014 authoritative status=active while a trip was held (age=${Math.round(decision.ageMs / 6e4)}min, trippedAt=${new Date(trip.trippedAt).toISOString()}); the pause this trip caused is no longer in force (ENG-7577)`
+  );
+  const marker = autoResumeMarkers.get(codeName);
+  const isSelfResume = marker !== void 0 && Date.now() - marker.autoResumedAt < AUTO_RESUME_SELF_WINDOW_MS;
+  if (!isSelfResume) autoResumeMarkers.delete(codeName);
+  state6 = {
+    ...state6,
+    circuitBreakerTrips: restartBreaker.serialize(),
+    circuitBreakerAutoResumes: Object.fromEntries(autoResumeMarkers.entries())
+  };
+  send({ type: "state-update", state: state6 });
+}
 var autoResumeMarkers = /* @__PURE__ */ new Map();
 var autoResumeInFlight = /* @__PURE__ */ new Set();
 var AUTO_RESUME_SELF_WINDOW_MS = 12e4;
@@ -8921,11 +9003,11 @@ function maybeAutoResume(agent) {
     now: Date.now()
   });
   if (decision.action === "skip") {
-    if ((decision.reason === "already-auto-resumed" || decision.reason === "not-quiet-yet") && trip) {
+    if ((decision.reason === "already-auto-resumed" || decision.reason === "not-quiet-yet" || decision.reason === "class-not-eligible") && trip) {
       const key = `${codeName}:${trip.trippedAt}`;
       if (autoResumeLoggedSkips.get(key) !== decision.reason) {
         autoResumeLoggedSkips.set(key, decision.reason);
-        const detail = decision.reason === "not-quiet-yet" ? ` (~${Math.ceil((decision.remainingMs ?? 0) / 6e4)}min remaining)` : " \u2014 credit spent for this cycle; operator resume re-arms it";
+        const detail = decision.reason === "not-quiet-yet" ? ` (~${Math.ceil((decision.remainingMs ?? 0) / 6e4)}min remaining)` : decision.reason === "class-not-eligible" ? ` class=${tripClass(trip)} \u2014 the blind quiet-timer only auto-recovers provisioning-reload trips; this one stays paused for an operator (ENG-7577)` : " \u2014 credit spent for this cycle; operator resume re-arms it";
         log(`[auto-resume] agent=${codeName} decision=skip reason=${decision.reason}${detail} (ENG-6088)`);
       }
     }
@@ -9299,6 +9381,7 @@ function transcriptAgeSecondsFor(codeName) {
   const sessionId = getSessionState(codeName)?.currentSessionId ?? null;
   return transcriptActivityAgeSeconds(getProjectDir2(codeName), sessionId, /* @__PURE__ */ new Date());
 }
+var restartGateDeferStreaks = /* @__PURE__ */ new Map();
 function restartGateFor(codeName, reason) {
   if (!isGateableRestartReason(reason)) return "bypass";
   const tzInputs = agentRestartTimezoneInputs.get(codeName);
@@ -9311,16 +9394,28 @@ function restartGateFor(codeName, reason) {
       tzInputs?.agentTimezoneExpiresAt ?? null
     )
   } : cachedMaintenanceWindow;
-  return decideRestartGate({
+  const streakKey = `${codeName}::${reason}`;
+  const priorStreak = restartGateDeferStreaks.get(streakKey) ?? 0;
+  const decision = decideRestartGate({
     window: perAgentWindow,
     // ENG-6491: channel adds (channel-set-change) skip the off-peak window but
     // still defer-until-idle — they come online as soon as the agent is quiet.
     windowExempt: isWindowExemptRestartReason(reason),
+    windowDeferStreak: priorStreak,
     paneLogAgeSeconds: paneLogAgeSecondsFor(codeName),
     transcriptAgeSeconds: transcriptAgeSecondsFor(codeName),
     inboundAgeSeconds: inboundAgeSecondsFor(codeName),
     now: /* @__PURE__ */ new Date()
   });
+  const nextStreak = nextWindowDeferStreak(priorStreak, decision);
+  if (nextStreak === 0) restartGateDeferStreaks.delete(streakKey);
+  else restartGateDeferStreaks.set(streakKey, nextStreak);
+  if (nextStreak === RESTART_WINDOW_DEFER_ESCALATE_AFTER && nextStreak !== priorStreak) {
+    log(
+      `[maintenance-window] '${codeName}': '${reason}' restart deferred ${nextStreak}\xD7 by the window \u2014 dropping the maintenance-window wait (ENG-8224). Still defer-until-idle: it will fire on the next quiet gap, never mid-turn.`
+    );
+  }
+  return decision;
 }
 function isHostBusyForForcedUpdate(opts) {
   const now = /* @__PURE__ */ new Date();
@@ -9787,7 +9882,7 @@ var agentRestartTimezoneInputs = /* @__PURE__ */ new Map();
 var lastVersionCheckAt = 0;
 var VERSION_CHECK_INTERVAL_MS = 5 * 60 * 1e3;
 var lastResponsivenessProbeAt = 0;
-var agtCliVersion = true ? "0.28.450" : "dev";
+var agtCliVersion = true ? "0.28.452" : "dev";
 function resolveBrewPath(execFileSync2) {
   try {
     const out = execFileSync2("which", ["brew"], { timeout: 5e3 }).toString().trim();
@@ -11778,6 +11873,7 @@ async function processAgent(agent, agentStates) {
   const previousKnownStatus = agentState.knownStatuses.get(agent.agent_id);
   agentState.agentDisplayNames.set(agent.code_name, agent.display_name);
   agentState.codeNameToAgentId.set(agent.code_name, agent.agent_id);
+  maybeClearStaleTrip(agent);
   maybeReportCurrentTrip(agent.code_name);
   if (agent.framework) {
     agentFrameworkCache.set(agent.code_name, agent.framework);
@@ -11920,6 +12016,7 @@ async function processAgent(agent, agentStates) {
     if (previousStatus === "paused" && agent.status === "active" && restartBreaker.isTripped(agent.code_name)) {
       restartBreaker.clear(agent.code_name);
       reportedTrips.delete(agent.code_name);
+      dependencyRecoveryLedger.clear(agent.code_name);
       log(`[circuit-breaker] Cleared trip for '${agent.code_name}' on operator resume (paused \u2192 active)`);
     }
     if (previousStatus === "paused" && agent.status === "active") {