@integrity-labs/agt-cli 0.28.758 → 0.28.759

package/dist/bin/agt.js

@@ -41,7 +41,7 @@ import {
   success,
   table,
   warn
-} from "../chunk-ZUCWB2AP.js";
+} from "../chunk-YAONHJW6.js";
 import {
   getProjectDir,
   isSessionResumeDisabled,
@@ -5467,7 +5467,7 @@ import { execFileSync, execSync } from "child_process";
 import { existsSync as existsSync11, realpathSync as realpathSync2 } from "fs";
 import chalk18 from "chalk";
 import ora16 from "ora";
-var cliVersion = true ? "0.28.758" : "dev";
+var cliVersion = true ? "0.28.759" : "dev";
 async function fetchLatestVersion() {
   const host2 = getHost();
   if (!host2) return null;
@@ -6658,7 +6658,7 @@ function handleError(err) {
 }
 
 // src/bin/agt.ts
-var cliVersion2 = true ? "0.28.758" : "dev";
+var cliVersion2 = true ? "0.28.759" : "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-ZUCWB2AP.js → chunk-YAONHJW6.js}

@@ -6436,7 +6436,7 @@ function exchangeFailureKind(err) {
 }
 
 // src/lib/api-client.ts
-var agtCliVersion = true ? "0.28.758" : "dev";
+var agtCliVersion = true ? "0.28.759" : "dev";
 var lastConfigHash = null;
 function setConfigHash(hash) {
   lastConfigHash = hash && hash.length > 0 ? hash : null;
@@ -10593,4 +10593,4 @@ export {
   managerInstallSystemUnitCommand,
   managerUninstallSystemUnitCommand
 };
-//# sourceMappingURL=chunk-ZUCWB2AP.js.map
+//# sourceMappingURL=chunk-YAONHJW6.js.map

package/dist/lib/manager-worker.js

@@ -58,7 +58,7 @@ import {
   safeWriteJsonAtomic,
   setConfigHash,
   tripClass
-} from "../chunk-ZUCWB2AP.js";
+} from "../chunk-YAONHJW6.js";
 import {
   getProjectDir as getProjectDir2,
   getReadyTasks,
@@ -393,6 +393,7 @@ function decideManagedMcpRestart(input) {
   const { prevStructureHash, structureHash, launchStructureHash, sessionHealthy, flapping } = input;
   if (prevStructureHash === void 0) return { kind: "skip-first-convergence" };
   if (structureHash === prevStructureHash) return { kind: "skip-no-change" };
+  if (input.deferringPrune === true) return { kind: "skip-prune-deferred" };
   if (!sessionHealthy) return { kind: "skip-unhealthy" };
   if (flapping) return { kind: "skip-flapping" };
   if (launchStructureHash != null && launchStructureHash === structureHash) {
@@ -6451,6 +6452,75 @@ function withScheduleLinkFooter(opts) {
   return appendScheduleLinkFooter(opts.body, url, opts.medium);
 }
 
+// src/lib/manager/managed-prune-streak.ts
+var DEFAULT_PRUNE_CONFIRM_CYCLES = 5;
+function pruneConfirmCycles(env = process.env) {
+  const raw = env["AUGMENTED_MANAGED_PRUNE_CONFIRM_CYCLES"];
+  if (raw === void 0) return DEFAULT_PRUNE_CONFIRM_CYCLES;
+  const trimmed = raw.trim();
+  if (!/^\d+$/.test(trimmed)) return DEFAULT_PRUNE_CONFIRM_CYCLES;
+  const n = Number(trimmed);
+  if (!Number.isSafeInteger(n) || n < 1) return DEFAULT_PRUNE_CONFIRM_CYCLES;
+  return n;
+}
+var ManagedPruneStreaks = class {
+  constructor(threshold = DEFAULT_PRUNE_CONFIRM_CYCLES) {
+    this.threshold = threshold;
+  }
+  streaks = /* @__PURE__ */ new Map();
+  /**
+   * Record one poll cycle for one agent.
+   *
+   * MUST BE CALLED ON EVERY CYCLE, including cycles where nothing changed —
+   * see the trap above. `declaredManagedKeys` is what the agent's config
+   * currently declares; `expectedKeys` is what the API says it should have.
+   *
+   * A key present in `expectedKeys` has its streak cleared IMMEDIATELY, which is
+   * what makes a toolkit's reappearance undo the wait rather than merely pause
+   * it.
+   *
+   * Returns a decision per absent key, in the order given, so the caller can log
+   * the deferrals as well as act on the matured ones.
+   */
+  observe(agentId, declaredManagedKeys, expectedKeys) {
+    let perAgent = this.streaks.get(agentId);
+    if (perAgent) {
+      for (const key of expectedKeys) perAgent.delete(key);
+    }
+    const decisions = [];
+    for (const key of declaredManagedKeys) {
+      if (expectedKeys.has(key)) continue;
+      if (!perAgent) {
+        perAgent = /* @__PURE__ */ new Map();
+        this.streaks.set(agentId, perAgent);
+      }
+      const streak = (perAgent.get(key) ?? 0) + 1;
+      perAgent.set(key, streak);
+      decisions.push({ serverKey: key, streak, mature: streak >= this.threshold });
+    }
+    if (perAgent) {
+      const declared = new Set(declaredManagedKeys);
+      for (const key of [...perAgent.keys()]) {
+        if (!declared.has(key)) perAgent.delete(key);
+      }
+      if (perAgent.size === 0) this.streaks.delete(agentId);
+    }
+    return decisions;
+  }
+  /** Current streak for one key. Zero when untracked. */
+  streakFor(agentId, serverKey) {
+    return this.streaks.get(agentId)?.get(serverKey) ?? 0;
+  }
+  /** Drop everything for an agent — used when an agent leaves the host. */
+  forget(agentId) {
+    this.streaks.delete(agentId);
+  }
+  /** Number of agents currently carrying at least one streak (tests/diagnostics). */
+  get trackedAgents() {
+    return this.streaks.size;
+  }
+};
+
 // src/lib/manager/agent-state.ts
 var agentState = {
   // ---------------------------------------------------------------------------
@@ -6558,7 +6628,19 @@ var agentState = {
    * it, so an unpublish during a restart window leaves the folder until the next
    * unpublish — acceptable for a platform-staff-rare action.
    */
-  knownGlobalSkillIds: /* @__PURE__ */ new Map()
+  knownGlobalSkillIds: /* @__PURE__ */ new Map(),
+  /**
+   * ENG-9708: consecutive cycles each managed MCP server key has been ABSENT
+   * from /host/managed-toolkits, per agent. A key must miss N consecutive polls
+   * before it is pruned, so one degraded-but-well-formed response can no longer
+   * destroy fleet state that does not self-heal (2026-08-31: 24 composio_*
+   * servers across 6+ agents, removed in under two minutes, recovered only by a
+   * manual bulk-reconnect).
+   *
+   * In-memory on purpose: a manager restart clears every streak, which delays a
+   * legitimate prune by a few minutes and can never cause a spurious one.
+   */
+  managedPruneStreaks: new ManagedPruneStreaks(pruneConfirmCycles())
 };
 function clearAgentState(agentId, codeName) {
   agentState.knownVersions.delete(agentId);
@@ -6577,6 +6659,7 @@ function clearAgentState(agentId, codeName) {
   agentState.knownMsTeamsBehaviourHashes.delete(agentId);
   agentState.knownSlackBehaviourHashes.delete(agentId);
   agentState.knownGlobalSkillIds.delete(agentId);
+  agentState.managedPruneStreaks.forget(agentId);
   agentState.agentDisplayNames.delete(codeName);
   agentState.codeNameToAgentId.delete(codeName);
   agentState.provisioningAttribution.delete(codeName);
@@ -13385,7 +13468,7 @@ var agentRestartTimezoneInputs = /* @__PURE__ */ new Map();
 var lastVersionCheckAt = 0;
 var VERSION_CHECK_INTERVAL_MS = 5 * 60 * 1e3;
 var lastResponsivenessProbeAt = 0;
-var agtCliVersion = true ? "0.28.758" : "dev";
+var agtCliVersion = true ? "0.28.759" : "dev";
 function resolveBrewPath(execFileSync3) {
   try {
     const out = execFileSync3("which", ["brew"], { timeout: 5e3 }).toString().trim();
@@ -17532,32 +17615,59 @@ async function processAgent(agent, agentStates, managedToolkits) {
           );
           recordConfigChurnEvent(agent.agent_id, agent.code_name, FLAP_CHANNEL_MANAGED_MCP, mcpMembership);
         }
+        const deferredPruneKeys = [];
+        if (frameworkAdapter.removeMcpServer && frameworkAdapter.readMcpServers) {
+          try {
+            const declaredManaged = Object.keys(
+              frameworkAdapter.readMcpServers(agent.code_name)
+            ).filter(isManagedMcpServerKey);
+            const decisions = agentState.managedPruneStreaks.observe(
+              agent.agent_id,
+              declaredManaged,
+              expectedServerIds
+            );
+            for (const d of decisions) {
+              if (!d.mature) {
+                deferredPruneKeys.push(d.serverKey);
+                log(
+                  `[managed-toolkit] ${agent.code_name}: '${d.serverKey}' absent for ${d.streak}/${pruneConfirmCycles()} cycles \u2014 deferring prune (ENG-9708)`
+                );
+                continue;
+              }
+              frameworkAdapter.removeMcpServer(agent.code_name, d.serverKey);
+              log(
+                `[managed-toolkit] Removed stale MCP server '${d.serverKey}' for '${agent.code_name}' after ${d.streak} consecutive absences (ENG-9708)`
+              );
+              managedPruneLedger.record(agent.code_name, d.serverKey);
+            }
+          } catch {
+          }
+        }
         if (mcpHash !== prevMcpHash) {
           for (const e of desiredEntries) {
             frameworkAdapter.writeMcpServer(agent.code_name, e.serverId, { url: e.url, headers: e.headers });
             const urlHash = createHash20("sha256").update(e.url).digest("hex").slice(0, 12);
             log(`[managed-toolkit] ${agent.code_name}: wrote '${e.name}' (serverId=${e.serverId}, url_hash=${urlHash})`);
           }
-          if (frameworkAdapter.removeMcpServer && frameworkAdapter.readMcpServers) {
-            try {
-              for (const key of Object.keys(frameworkAdapter.readMcpServers(agent.code_name))) {
-                if (isManagedMcpServerKey(key) && !expectedServerIds.has(key)) {
-                  frameworkAdapter.removeMcpServer(agent.code_name, key);
-                  log(`[managed-toolkit] Removed stale MCP server '${key}' for '${agent.code_name}'`);
-                  managedPruneLedger.record(agent.code_name, key);
-                }
-              }
-            } catch {
-            }
+          if (deferredPruneKeys.length === 0) {
+            agentState.knownManagedMcpHashes.set(agent.agent_id, mcpHash);
+            agentState.knownManagedMcpStructure.set(agent.agent_id, structureHash);
+          } else {
+            log(
+              `[managed-toolkit] ${agent.code_name}: holding the managed-MCP baseline and session restart while ${deferredPruneKeys.length} prune(s) await confirmation (${deferredPruneKeys.join(", ")}) \u2014 ENG-9708`
+            );
           }
-          agentState.knownManagedMcpHashes.set(agent.agent_id, mcpHash);
-          agentState.knownManagedMcpStructure.set(agent.agent_id, structureHash);
           const mcpRestartDecision = fwForMcp === "claude-code" && prevStructureHash !== void 0 && structureHash !== prevStructureHash ? decideManagedMcpRestart({
             prevStructureHash,
             structureHash,
             launchStructureHash: sessionLaunchManagedStructure.get(agent.code_name),
             sessionHealthy: isSessionHealthy(agent.code_name),
-            flapping: mcpFlap.flapping
+            flapping: mcpFlap.flapping,
+            // ENG-9708: an unconfirmed prune holds the restart too. The
+            // structural change it would announce ("servers were removed")
+            // has not been applied to the file yet, so restarting now
+            // would churn the session and tell the agent something untrue.
+            deferringPrune: deferredPruneKeys.length > 0
           }) : null;
           if (mcpRestartDecision?.kind === "suppress-already-live") {
             log(