@integrity-labs/agt-cli 0.28.923 → 0.28.924

package/dist/bin/agt.js

@@ -40,7 +40,7 @@ import {
   success,
   table,
   warn
-} from "../chunk-D3DVYP4X.js";
+} from "../chunk-FAG5UIWZ.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 ? "0.28.923" : "dev";
+var cliVersion = true ? "0.28.924" : "dev";
 async function fetchLatestVersion() {
   const host2 = getHost();
   if (!host2) return null;
@@ -6682,7 +6682,7 @@ function handleError(err) {
 }
 
 // src/bin/agt.ts
-var cliVersion2 = true ? "0.28.923" : "dev";
+var cliVersion2 = true ? "0.28.924" : "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-D3DVYP4X.js → chunk-FAG5UIWZ.js}

@@ -6942,7 +6942,7 @@ function exchangeFailureKind(err) {
 }
 
 // src/lib/api-client.ts
-var agtCliVersion = true ? "0.28.923" : "dev";
+var agtCliVersion = true ? "0.28.924" : "dev";
 var lastConfigHash = null;
 function setConfigHash(hash) {
   lastConfigHash = hash && hash.length > 0 ? hash : null;
@@ -11552,4 +11552,4 @@ export {
   managerInstallSystemUnitCommand,
   managerUninstallSystemUnitCommand
 };
-//# sourceMappingURL=chunk-D3DVYP4X.js.map
+//# sourceMappingURL=chunk-FAG5UIWZ.js.map

package/dist/lib/manager-worker.js

@@ -63,7 +63,7 @@ import {
   safeWriteJsonAtomic,
   setConfigHash,
   tripClass
-} from "../chunk-D3DVYP4X.js";
+} from "../chunk-FAG5UIWZ.js";
 import {
   getProjectDir as getProjectDir2,
   getReadyTasks,
@@ -3536,6 +3536,51 @@ async function maybeReconcileWorkflowRunTokens(args) {
 // src/lib/conversation-evaluator.ts
 import { readdirSync as readdirSync3, readFileSync as readFileSync12, statSync as statSync3 } from "fs";
 import { join as join17 } from "path";
+
+// src/lib/detect-meta-egress.ts
+var REPOST_PREAMBLE_WINDOW = 200;
+var REPOST_ANNOUNCE = /\b(?:re-?(?:post|send)(?:ing)?\s+(?:it|this|that)\b|(?:posting|sending)\s+(?:it|this|that)\s+(?:now|again)\b|re-?posting\s+now\b)/i;
+var REPOST_FAILURE = /\b(?:it|this|that)\s+(?:did(?:n['’]t| not)|could(?:n['’]t| not)|wasn['’]t|failed to)\s+(?:parse|send|post|go through|land|deliver|come through)\b|\b(?:did(?:n['’]t| not)|could(?:n['’]t| not)|failed to)\s+(?:send|post|deliver)\s+(?:it|this|that)\b/i;
+var REPOST_PRIOR = /\b(?:drafted|dialled|dialed|typed|composed|wrote)\s+(?:this|that|it)\s+(?:a moment ago|moments ago|just now|earlier)\b/i;
+function hasRepostPreamble(message) {
+  const head = message.slice(0, REPOST_PREAMBLE_WINDOW);
+  return REPOST_ANNOUNCE.test(head) && (REPOST_FAILURE.test(head) || REPOST_PRIOR.test(head));
+}
+var STATUS_NARRATION = [
+  // "Assessment delivered to Samuel:" - a ledger line: the noun OPENS a sentence
+  // (or parenthetical) and the recipient is a capitalised name. Case-sensitive
+  // on purpose, so "Summary delivered to the board" and "Your answer delivered
+  // to the customer" are not matched. \p{Lu} (not [A-Z]) so a non-ASCII
+  // recipient ("Élodie") is still a capitalised name.
+  new RegExp("(?:^|[.!?]\\s+|\\(\\s*)(?:[Aa]ssessment|[Ss]ummary|[Rr]eport|[Aa]nalysis|[Rr]esponse|[Rr]eply|[Uu]pdate|[Aa]nswer)\\s+delivered\\s+to\\s+\\p{Lu}", "u"),
+  // third-person "on his/her/their go" (waiting on the user's go-ahead, narrated)
+  /\bon (?:his|her|their) go\b(?![-\w])/i
+];
+var LOOP_TICK = [
+  // "Standing down - no change." / "Standing down, no action needed"
+  /\bstanding down\s*[—–,:;.-]?\s*no (?:change|changes|action needed|new work)\b/i,
+  // A message that is ONLY "Standing down."
+  /^standing down[.!]?$/i,
+  // "All clear, no pending work."
+  /\ball clear\s*[,—–:;-]?\s*no pending work\b/i
+];
+function detectMetaEgress(agentMessages) {
+  const kinds = /* @__PURE__ */ new Set();
+  for (const raw of agentMessages) {
+    if (!raw) continue;
+    const m = raw.trim();
+    if (!m) continue;
+    if (hasRepostPreamble(m)) kinds.add("repost_preamble");
+    if (STATUS_NARRATION.some((re) => re.test(m))) kinds.add("status_narration");
+    if (LOOP_TICK.some((re) => re.test(m))) kinds.add("loop_tick");
+  }
+  return {
+    meta_leak: kinds.size > 0,
+    meta_leak_kinds: [...kinds].sort()
+  };
+}
+
+// src/lib/conversation-evaluator.ts
 var MIN_CHECK_INTERVAL_MS4 = 5 * 6e4;
 var TRANSCRIPT_MTIME_WINDOW_MS3 = 7 * 24 * 60 * 60 * 1e3;
 var WINDOW_PAD_MS = 5 * 6e4;
@@ -3908,6 +3953,9 @@ async function maybeEvaluateConversations(args) {
       continue;
     }
     backendSucceeded = true;
+    const flags = detectMetaEgress(
+      turns.filter((t) => t.role === "assistant").map((t) => t.text)
+    );
     try {
       await api2.post("/host/conversations/evaluation", {
         agent_id: agentId,
@@ -3916,7 +3964,8 @@ async function maybeEvaluateConversations(args) {
         verdict: verdict.verdict,
         summary: verdict.summary,
         failure_category: verdict.failureCategory,
-        model: backend.model
+        model: backend.model,
+        eval_flags: flags
       });
       log2(
         `[conversation-eval] ${codeName}: ${conv.conversation_id.slice(0, 8)} \u2192 ${verdict.verdict} (${verdict.score})`
@@ -13823,7 +13872,7 @@ var agentRestartTimezoneInputs = /* @__PURE__ */ new Map();
 var lastVersionCheckAt = 0;
 var VERSION_CHECK_INTERVAL_MS = 5 * 60 * 1e3;
 var lastResponsivenessProbeAt = 0;
-var agtCliVersion = true ? "0.28.923" : "dev";
+var agtCliVersion = true ? "0.28.924" : "dev";
 function resolveBrewPath(execFileSync2) {
   try {
     const out = execFileSync2("which", ["brew"], { timeout: 5e3 }).toString().trim();