@@ -902,7 +902,7 @@ npm run promote -- --pipeline <name> --from <source> --to <target> --apply # For
 npm run validate -- <org>                          # Lint resources locally (fails fast on schema drift)
 npm run audit -- <org>                             # Read-only drift detector: orphan YAML, state ghosts, content-identical clusters, sibling base-slugs, dashboard orphans, inline model.tools. Exit 1 on findings.
 npm run audit -- <org> --type assistants           # Scope audit to a single resource type
-npm run sim -- <org> --suite <name> --target <name>  # Run a simulation suite against an assistant/squad
+npm run sim -- <org> --suite <name> --target <name>  # Run a simulation suite against an assistant/squad (exit 0 pass, 1 fail, 3 incomplete; --timeout <min>)
 npm run rollback -- <org> --to <ISO-timestamp>     # Re-apply a snapshot taken before a push
 npm run rollback -- <org> --list                   # List available snapshots
 
 
@@ -103,7 +103,7 @@ Every command works in two modes:
 | `npm run cleanup` | ✅ | `npm run cleanup -- <org> [--force --confirm <org>]` | Inspect (default) or delete orphaned remote resources. Destructive run requires `--confirm <org>`. |
 | `npm run rollback` | — | `npm run rollback -- <org> --list` or `--to <ISO>` | Restore from a snapshot in `.vapi-state.<org>.snapshots/` (one is written before every push/apply). |
 | `npm run call` | ✅ | `npm run call -- <org> -a <name>` or `-s <squad>` | Start an interactive WebSocket call against an assistant or squad. |
-| `npm run sim` | — | `npm run sim -- <org> --suite <name> --target <name>` | Run a simulation suite (or specific simulations) against a deployed assistant/squad. |
+| `npm run sim` | — | `npm run sim -- <org> --suite <name> --target <name> [--timeout <min>]` | Run a simulation suite (or specific simulations) against a deployed assistant/squad. Prints the run link; exits 0 passed, 1 failed, 3 incomplete (timeout, Ctrl-C, missing results). |
 | `npm run migrate` | — | `npm run migrate` | One-time, all orgs at once: slim legacy state files to pure `name → uuid` and seed the per-developer `.vapi-state-hash/` baseline store from the old hashes. Required once after upgrading to the hash-store engine — `pull`/`push`/`apply` refuse legacy-shaped state until it runs. Idempotent. |
 | `npm run build` | — | — | Type-check the codebase (`tsc --noEmit`). |
 | `npm test` | — | — | Run regression tests (`node:test`). |
 
@@ -471,6 +471,23 @@ This is the unified executor. A "run" is a batch — it expands into many `runIt
 }
 ```
 
+The **create** response (`CreateSimulationRunResponse`) also carries two fields nothing else returns:
+
+- `url` — the dashboard link to this run. `GET /eval/simulation/run/:id` does **not** return it, so read it from the create response.
+- `simulationRunItemIds` — one id per item queued (simulations × iterations), i.e. how many items the run should end with.
+
+A run has **no `results` field**. Pass/fail lives in `itemCounts` and on the items themselves (`GET …/item`, below). Creating a run queues paid work before the response returns, so a 5xx on create may still have started it — don't blindly retry the POST.
+
+### How `npm run sim` decides pass/fail
+
+`src/sim-result.ts` (`simRunVerdict`) reports **passed** only when all of these hold; otherwise **failed** (any failed item) or **incomplete**:
+
+- the run `ended` and has `itemCounts`, with `total` equal to the number of items created (`simulationRunItemIds`) and greater than 0;
+- nothing is `queued`, `running` or `canceled`, and `passed === total`;
+- every item was fetched, is `passed`, and has at least one **required** evaluation that wasn't skipped (the all-skipped trap from the chat-mode gotcha above).
+
+Exit codes: 0 passed, 1 failed, 2 usage error, 3 incomplete (timeout, Ctrl-C, canceled items, results that never arrived). On timeout or Ctrl-C the run is canceled.
+
 ### List runs — `GET /eval/simulation/run`
 
 Query params:
@@ -486,7 +503,7 @@ Query params:
 ### Get / Cancel run
 
 - `GET /eval/simulation/run/:id` → `SimulationRun`
-- `PATCH /eval/simulation/run/:id` → cancels the run **and** all its queued items. No body required.
+- `PATCH /eval/simulation/run/:id` → cancels the run **and** all its queued items. No body required. Returns 400 `Run has already ended` for an ended run and 409 when a concurrent cancel wins; both mean "nothing left to cancel".
 
 ---
 
@@ -498,6 +515,8 @@ Run items are system-managed — there's no create/update API for users; they're
 
 Query params: `limit`, `page`, `simulationId`, `runId`, `status` (`queued` | `running` | `evaluating` | `passed` | `failed` | `canceled`).
 
+Response shape depends on the query: **with** `limit` or `page` it's paginated (`{ results, metadata: { totalItems, itemsPerPage, currentPage } }`, `limit` up to 1000); **without** them it's a bare array. Pages are ordered only by creation time and a run's items share it, so OFFSET pages can overlap — dedupe by `id`. Items can also lag the run: a run can be `ended` while an item is `passed` but its `results` aren't written yet.
+
 ### Get a run item — `GET /eval/simulation/run/:id/item/:itemId`
 
 **Response** — `SimulationRunItem` is rich; the highlights:
 
@@ -84,6 +84,7 @@ you which stack PR closes the row.**
 | 30  | Tool-linking pass could PATCH a raw assistant slug          | Mid-push 400 naming the wrong resource              | None       | RESOLVED 2026-08-03 (#51)                       |
 | 31  | Unresolved references handled 3 inconsistent ways, no dangling-ref check | Same authoring mistake, three different failure modes | None | Open                                            |
 | 32  | Test suite never ran in CI; 20 tests rotted after the hash store | Regression guards for #22/#23 silently stopped running | None | RESOLVED 2026-09-30 (#56)                      |
+| 33  | `npm run sim` reported every run as passed                | A failing suite exited 0 — false green             | None       | RESOLVED 2026-10-01                             |
 
 **Active backlog after cleanup:** `#2`, `#6`, `#8`, `#12`, `#20`, `#24–#26`, `#31`, and the open remainder of `#27` (wiring the listing-completeness verdict into push/delete/audit, and moving `cleanup.ts` onto the shared pager). Resolved entries stay in this file as historical incident notes per the maintenance directive; stale superseded backlog rows are not duplicated.
 
@@ -1721,6 +1722,57 @@ fields.
 
 ---
 
+## 33. `npm run sim` reported every run as passed
+
+**[RESOLVED 2026-10-01]**
+
+**Discovered:** 2026-10-01, while designing simulation PR checks (TEST-141).
+
+### Problem
+
+`npm run sim` scored a run by reading `results[]` on the run and counting
+`status === "pass"`. The simulation-run API has no `results` field and
+item statuses are `passed` / `failed`, so every watched run summarised
+as 0 pass / 0 fail and the command exited 0 — including runs whose
+simulations failed.
+
+### Current behavior (Verified, before the fix)
+
+- `src/sim.ts` `runSimulation` computed pass/fail from `last.results`,
+  which `GET /eval/simulation/run/:id` never returns; pass/fail is in
+  `itemCounts` and on the run items (`GET /eval/simulation/run/:id/item`).
+- `src/sim-cmd.ts` exited 1 only when `fail > 0`, which could never happen.
+- Polling also stopped on statuses that don't exist (`failed`,
+  `completed`), never canceled a timed-out run, and didn't print the run
+  link (only the create response carries `url`).
+
+### Risk
+
+Anyone gating on `npm run sim` (locally or in CI) got a green result for a
+failing suite.
+
+### Current mitigation
+
+None needed once the fix below lands.
+
+### Possible fix (landed)
+
+- `src/sim-result.ts` `simRunVerdict`: passed only when the run ended,
+  every expected item exists, passed, and had a required evaluation that was
+  actually scored; otherwise failed or incomplete with a reason.
+- `src/sim.ts` reads items (paginated or bare-array, deduped by id),
+  waits for late item results, prints the run link and failing judges, and
+  cancels the run on `--timeout` (default 20 min) or Ctrl-C.
+- `src/vapi-client.ts`: a config-free client that never retries run
+  creation on a 5xx (the run may already be queued).
+- Exit codes: 0 passed, 1 failed, 2 usage, 3 incomplete.
+
+### Status
+
+**RESOLVED 2026-10-01.**
+
+---
+
 ## Out of scope (intentionally not improvements)
 
 - **State file is identity-only and not git-ignored.** It's intentionally
 
@@ -2,7 +2,12 @@
 //
 // Wraps `POST /eval/simulation/run` (the simulations API). See AGENTS.md for
 // usage.
+//
+// Exit codes: 0 passed, 1 failed, 2 usage/config error, 3 incomplete
+// (timed out, interrupted, canceled, or results that never fully arrived).
 
+import { resolve } from "node:path";
+import { fileURLToPath } from "node:url";
 import {
   formatSummary,
   loadEnvFile,
@@ -12,27 +17,28 @@ import {
   runSimulation,
 } from "./sim.ts";
 
-function printUsage(): void {
-  console.error(
-    [
-      "Usage:",
-      "  npm run sim -- <org> --suite <suite-name> --target <assistant-or-squad-name>",
-      "  npm run sim -- <org> --simulations <name1>,<name2> --target <assistant-name>",
-      "",
-      "Options:",
-      "  --suite <name>         Run an entire simulation suite by local resource name",
-      "  --simulations <list>   Run one or more simulations by comma-separated local names",
-      "  --target <name>        Local assistant or squad name (resolves to UUID via state)",
-      "  --transport voice|chat Transport (default: voice; chat is faster/cheaper)",
-      "  --iterations N         Override default iteration count",
-      "  --watch                Tail status until completion (default: on)",
-      "",
-      "Examples:",
-      "  npm run sim -- my-org --suite booking-tests --target intake-agent",
-      "  npm run sim -- my-org --simulations happy-path,edge-case --target main-agent --transport chat",
-    ].join("\n"),
-  );
-}
+const USAGE = [
+  "Usage:",
+  "  npm run sim -- <org> --suite <suite-name> --target <assistant-or-squad-name>",
+  "  npm run sim -- <org> --simulations <name1>,<name2> --target <assistant-name>",
+  "",
+  "Options:",
+  "  --suite <name>         Run an entire simulation suite by local resource name",
+  "  --simulations <list>   Run one or more simulations by comma-separated local names",
+  "  --target <name>        Local assistant or squad name (resolves to UUID via state)",
+  "  --transport voice|chat Transport (default: voice; chat is faster/cheaper)",
+  "  --iterations N         Override default iteration count",
+  "  --timeout <minutes>    Give up and cancel the run after this long (default: 20)",
+  "  --no-watch             Start the run, print its link, and exit 0 without a verdict",
+  "",
+  "Exit codes: 0 passed, 1 failed, 2 usage error, 3 incomplete (timeout, interrupt, missing results)",
+  "",
+  "Examples:",
+  "  npm run sim -- my-org --suite booking-tests --target intake-agent",
+  "  npm run sim -- my-org --simulations happy-path,edge-case --target main-agent --transport chat",
+].join("\n");
+
+class UsageError extends Error {}
 
 interface ParsedArgs {
   env: string;
@@ -42,71 +48,83 @@ interface ParsedArgs {
   squad?: string;
   transport?: "voice" | "chat";
   iterations?: number;
+  timeoutMinutes?: number;
   watch: boolean;
+  help: boolean;
 }
 
-function parseArgs(): ParsedArgs {
-  const args = process.argv.slice(2);
+function argsParse(args: string[]): ParsedArgs {
   const env = args[0];
-  if (!env) {
-    printUsage();
-    process.exit(1);
+  if (!env || env === "--help" || env === "-h") {
+    return { env: env ?? "", watch: true, help: true };
   }
   const SLUG_RE = /^[a-z0-9]([a-z0-9-]*[a-z0-9])?$/;
-  if (!SLUG_RE.test(env)) {
-    console.error(`❌ Invalid org name: ${env}`);
-    process.exit(1);
-  }
-
-  const parsed: ParsedArgs = { env, watch: true };
+  if (!SLUG_RE.test(env)) throw new UsageError(`Invalid org name: ${env}`);
 
+  const parsed: ParsedArgs = { env, watch: true, help: false };
   for (let i = 1; i < args.length; i++) {
     const arg = args[i];
     if (arg === "--suite") parsed.suite = args[++i];
     else if (arg === "--simulations") parsed.simulations = args[++i];
     else if (arg === "--target") {
       // We don't know yet whether target is an assistant or squad — defer
-      // resolution to the state lookup. Try assistant first; resolveTarget()
-      // accepts either argument key, so we set the candidate in `assistant`
-      // and let `resolveTarget` fall through to `squad` if not found.
-      // For clarity, we accept --assistant / --squad as explicit alternatives.
+      // resolution to the state lookup (see simCommandRun).
       parsed.assistant = args[++i];
     } else if (arg === "--assistant") parsed.assistant = args[++i];
     else if (arg === "--squad") parsed.squad = args[++i];
     else if (arg === "--transport") {
       const v = args[++i];
-      if (v === "voice" || v === "chat") parsed.transport = v;
-      else {
-        console.error(`❌ --transport must be "voice" or "chat" (got "${v}")`);
-        process.exit(1);
+      if (v !== "voice" && v !== "chat") {
+        throw new UsageError(
+          `--transport must be "voice" or "chat" (got "${v}")`,
+        );
       }
+      parsed.transport = v;
     } else if (arg === "--iterations") {
-      parsed.iterations = parseInt(args[++i] ?? "", 10);
+      parsed.iterations = Number.parseInt(args[++i] ?? "", 10);
       if (Number.isNaN(parsed.iterations)) {
-        console.error("❌ --iterations requires a number");
-        process.exit(1);
+        throw new UsageError("--iterations requires a number");
+      }
+    } else if (arg === "--timeout") {
+      parsed.timeoutMinutes = Number(args[++i]);
+      if (
+        !Number.isFinite(parsed.timeoutMinutes) ||
+        parsed.timeoutMinutes <= 0
+      ) {
+        throw new UsageError("--timeout requires a positive number of minutes");
       }
     } else if (arg === "--no-watch") parsed.watch = false;
     else if (arg === "--watch") parsed.watch = true;
-    else if (arg === "--help" || arg === "-h") {
-      printUsage();
-      process.exit(0);
-    }
+    else if (arg === "--help" || arg === "-h") parsed.help = true;
+    else throw new UsageError(`Unknown argument: ${arg}`);
   }
-
   return parsed;
 }
 
-async function main(): Promise<void> {
-  const args = parseArgs();
-  const cfg = loadEnvFile(args.env);
-  const state = loadStateFile(args.env);
+export async function simCommandRun(
+  args = process.argv.slice(2),
+): Promise<number> {
+  let parsed: ParsedArgs;
+  try {
+    parsed = argsParse(args);
+  } catch (error) {
+    if (!(error instanceof UsageError)) throw error;
+    console.error(`❌ ${error.message}\n\n${USAGE}`);
+    return 2;
+  }
+  if (parsed.help) {
+    console.log(USAGE);
+    return parsed.env ? 0 : 2;
+  }
+
+  const cfg = loadEnvFile(parsed.env);
+  const state = loadStateFile(parsed.env);
 
   // Disambiguate --target: if the bare value matches a squad name in state
   // and not an assistant, treat it as a squad. Explicit --assistant / --squad
   // override the heuristic.
-  let assistant = args.assistant;
-  let squad = args.squad;
+  let assistant = parsed.assistant;
+  let squad = parsed.squad;
   if (assistant && !squad) {
     const isSquad =
       typeof state.squads[assistant] !== "undefined" &&
@@ -120,39 +138,71 @@ async function main(): Promise<void> {
   console.log(
     "═══════════════════════════════════════════════════════════════",
   );
-  console.log(`🧪 Vapi GitOps Sim Runner — Environment: ${args.env}`);
+  console.log(`🧪 Vapi GitOps Sim Runner — Environment: ${parsed.env}`);
   console.log(`   API: ${cfg.baseUrl}`);
   console.log(
     "═══════════════════════════════════════════════════════════════\n",
   );
 
   const selection = resolveSelection(state, {
-    suite: args.suite,
-    simulations: args.simulations,
+    suite: parsed.suite,
+    simulations: parsed.simulations,
   });
   const target = resolveTarget(state, { assistant, squad });
 
-  const summary = await runSimulation(cfg, selection, target, {
-    watch: args.watch,
-    iterations: args.iterations,
-    transport: args.transport,
-  });
+  // Ctrl-C / SIGTERM cancel the run instead of leaving it running unwatched.
+  const controller = new AbortController();
+  const stop = () => controller.abort();
+  process.on("SIGINT", stop);
+  process.on("SIGTERM", stop);
+  let summary;
+  try {
+    summary = await runSimulation(cfg, selection, target, {
+      watch: parsed.watch,
+      iterations: parsed.iterations,
+      transport: parsed.transport,
+      timeoutMs:
+        parsed.timeoutMinutes === undefined
+          ? undefined
+          : parsed.timeoutMinutes * 60_000,
+      signal: controller.signal,
+    });
+  } finally {
+    process.off("SIGINT", stop);
+    process.off("SIGTERM", stop);
+  }
 
   console.log(`\n${formatSummary(summary)}\n`);
-
-  if (summary.fail > 0) {
-    console.error(
-      `❌ Simulation run failed (${summary.fail} fail / ${summary.pass} pass)`,
-    );
-    process.exit(1);
+  if (!summary.verdict) {
+    console.log("▶️  Run started (not watched).");
+    return 0;
+  }
+  if (summary.verdict.status === "passed") {
+    console.log("✅ Simulation run passed.");
+    return 0;
   }
-  console.log("✅ Simulation run passed.");
+  if (summary.verdict.status === "failed") {
+    console.error(`❌ Simulation run failed: ${summary.verdict.reason}`);
+    return 1;
+  }
+  console.error(`⚠️  Simulation run incomplete: ${summary.verdict.reason}`);
+  return 3;
 }
 
-main().catch((error) => {
-  console.error(
-    "\n❌ Sim failed:",
-    error instanceof Error ? error.message : error,
+const isMainModule =
+  process.argv[1] !== undefined &&
+  resolve(process.argv[1]) === fileURLToPath(import.meta.url);
+if (isMainModule) {
+  simCommandRun().then(
+    (code) => {
+      process.exitCode = code;
+    },
+    (error: unknown) => {
+      console.error(
+        "\n❌ Sim failed:",
+        error instanceof Error ? error.message : error,
+      );
+      process.exitCode = 2;
+    },
   );
-  process.exit(1);
-});
+}