@vellumai/assistant 0.6.0 → 0.6.2

package/src/__tests__/profiler-run-store.test.ts

@@ -0,0 +1,441 @@
+/**
+ * Tests for the profiler run store: manifest management, retention sweep,
+ * active-run protection, oldest-first pruning, max-run-count pruning,
+ * active-run-over-budget signaling, and idempotent rescans.
+ */
+import {
+  existsSync,
+  mkdirSync,
+  readFileSync,
+  rmSync,
+  writeFileSync,
+} from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import { afterEach, beforeEach, describe, expect, test } from "bun:test";
+
+import type { ProfilerRunManifest } from "../daemon/profiler-run-store.js";
+import { rescanRuns, runProfilerSweep } from "../daemon/profiler-run-store.js";
+
+// ── Test scaffolding ────────────────────────────────────────────────────
+
+let testDir: string;
+let runsDir: string;
+let origEnv: Record<string, string | undefined>;
+
+/**
+ * Create a fake profiler run directory with some payload files.
+ */
+function createRun(
+  runId: string,
+  opts?: {
+    sizeBytes?: number;
+    manifest?: Partial<ProfilerRunManifest>;
+  },
+): string {
+  const dir = join(runsDir, runId);
+  mkdirSync(dir, { recursive: true });
+
+  // Write a payload file of the requested size
+  const size = opts?.sizeBytes ?? 1024;
+  writeFileSync(join(dir, "profile.cpuprofile"), Buffer.alloc(size));
+
+  // Optionally write a pre-existing manifest
+  if (opts?.manifest) {
+    const m: ProfilerRunManifest = {
+      runId,
+      status: opts.manifest.status ?? "completed",
+      createdAt: opts.manifest.createdAt ?? new Date().toISOString(),
+      updatedAt: opts.manifest.updatedAt ?? new Date().toISOString(),
+      totalBytes: opts.manifest.totalBytes ?? size,
+    };
+    writeFileSync(join(dir, "manifest.json"), JSON.stringify(m, null, 2));
+  }
+
+  return dir;
+}
+
+function readManifestFromDisk(runId: string): ProfilerRunManifest | null {
+  const manifestPath = join(runsDir, runId, "manifest.json");
+  try {
+    return JSON.parse(readFileSync(manifestPath, "utf-8"));
+  } catch {
+    return null;
+  }
+}
+
+beforeEach(() => {
+  testDir = join(
+    tmpdir(),
+    `vellum-profiler-test-${Date.now()}-${Math.random().toString(36).slice(2)}`,
+  );
+  runsDir = join(testDir, "data", "profiler", "runs");
+  mkdirSync(runsDir, { recursive: true });
+
+  // Save and override env
+  origEnv = {
+    VELLUM_WORKSPACE_DIR: process.env.VELLUM_WORKSPACE_DIR,
+    VELLUM_PROFILER_RUN_ID: process.env.VELLUM_PROFILER_RUN_ID,
+    VELLUM_PROFILER_MAX_BYTES: process.env.VELLUM_PROFILER_MAX_BYTES,
+    VELLUM_PROFILER_MAX_RUNS: process.env.VELLUM_PROFILER_MAX_RUNS,
+    VELLUM_PROFILER_MIN_FREE_MB: process.env.VELLUM_PROFILER_MIN_FREE_MB,
+  };
+
+  // Point workspace dir to our temp directory
+  process.env.VELLUM_WORKSPACE_DIR = testDir;
+
+  // Clear profiler env vars
+  delete process.env.VELLUM_PROFILER_RUN_ID;
+  delete process.env.VELLUM_PROFILER_MAX_BYTES;
+  delete process.env.VELLUM_PROFILER_MAX_RUNS;
+  delete process.env.VELLUM_PROFILER_MIN_FREE_MB;
+});
+
+afterEach(() => {
+  // Restore env
+  for (const [key, value] of Object.entries(origEnv)) {
+    if (value === undefined) {
+      delete process.env[key];
+    } else {
+      process.env[key] = value;
+    }
+  }
+
+  // Clean up temp directory
+  if (existsSync(testDir)) {
+    rmSync(testDir, { recursive: true, force: true });
+  }
+});
+
+// ── Tests ───────────────────────────────────────────────────────────────
+
+describe("Profiler run store", () => {
+  describe("rescanRuns", () => {
+    test("returns empty array when no runs directory exists", () => {
+      // Remove the runs directory
+      rmSync(runsDir, { recursive: true, force: true });
+      const manifests = rescanRuns();
+      expect(manifests).toEqual([]);
+    });
+
+    test("returns empty array when runs directory is empty", () => {
+      const manifests = rescanRuns();
+      expect(manifests).toEqual([]);
+    });
+
+    test("creates manifests for run directories without existing manifests", () => {
+      createRun("run-001", { sizeBytes: 2048 });
+      createRun("run-002", { sizeBytes: 4096 });
+
+      const manifests = rescanRuns();
+      expect(manifests).toHaveLength(2);
+
+      const run1 = manifests.find((m) => m.runId === "run-001");
+      expect(run1).toBeDefined();
+      expect(run1!.status).toBe("completed");
+      // totalBytes includes manifest.json that rescan just wrote
+      expect(run1!.totalBytes).toBeGreaterThanOrEqual(2048);
+
+      const run2 = manifests.find((m) => m.runId === "run-002");
+      expect(run2).toBeDefined();
+      expect(run2!.status).toBe("completed");
+      expect(run2!.totalBytes).toBeGreaterThanOrEqual(4096);
+    });
+
+    test("marks the active run correctly", () => {
+      process.env.VELLUM_PROFILER_RUN_ID = "active-run";
+      createRun("active-run", { sizeBytes: 1024 });
+      createRun("old-run", { sizeBytes: 1024 });
+
+      const manifests = rescanRuns();
+      const active = manifests.find((m) => m.runId === "active-run");
+      const old = manifests.find((m) => m.runId === "old-run");
+
+      expect(active!.status).toBe("active");
+      expect(old!.status).toBe("completed");
+    });
+
+    test("transitions previously-active run to completed when no longer active", () => {
+      // Create a run with an "active" manifest
+      createRun("old-active", {
+        sizeBytes: 1024,
+        manifest: { status: "active", createdAt: "2025-01-01T00:00:00Z" },
+      });
+
+      // No VELLUM_PROFILER_RUN_ID set, so nothing is active
+      const manifests = rescanRuns();
+      const run = manifests.find((m) => m.runId === "old-active");
+
+      expect(run!.status).toBe("completed");
+
+      // Verify it was persisted to disk
+      const onDisk = readManifestFromDisk("old-active");
+      expect(onDisk!.status).toBe("completed");
+    });
+
+    test("is idempotent — repeated calls after initial scan produce the same result", () => {
+      createRun("run-a", { sizeBytes: 1024 });
+      process.env.VELLUM_PROFILER_RUN_ID = "run-a";
+
+      // First call writes the manifest, which changes totalBytes
+      rescanRuns();
+      // Second and third calls should be stable
+      const second = rescanRuns();
+      const third = rescanRuns();
+
+      expect(second).toHaveLength(1);
+      expect(third).toHaveLength(1);
+      expect(second[0]!.runId).toBe(third[0]!.runId);
+      expect(second[0]!.status).toBe(third[0]!.status);
+      expect(second[0]!.totalBytes).toBe(third[0]!.totalBytes);
+    });
+
+    test("preserves createdAt from existing manifest", () => {
+      const originalCreatedAt = "2024-06-15T12:00:00Z";
+      createRun("preserved-run", {
+        sizeBytes: 1024,
+        manifest: {
+          status: "completed",
+          createdAt: originalCreatedAt,
+        },
+      });
+
+      const manifests = rescanRuns();
+      const run = manifests.find((m) => m.runId === "preserved-run");
+      expect(run!.createdAt).toBe(originalCreatedAt);
+    });
+  });
+
+  describe("runProfilerSweep", () => {
+    test("no-ops when no runs exist", () => {
+      const result = runProfilerSweep();
+      expect(result.prunedCount).toBe(0);
+      expect(result.freedBytes).toBe(0);
+      expect(result.activeRunOverBudget).toBe(false);
+      expect(result.remainingRuns).toBe(0);
+    });
+
+    test("does not prune when under all budgets", () => {
+      process.env.VELLUM_PROFILER_MAX_BYTES = "1000000"; // 1 MB
+      process.env.VELLUM_PROFILER_MAX_RUNS = "10";
+
+      createRun("run-1", { sizeBytes: 1024 });
+      createRun("run-2", { sizeBytes: 1024 });
+
+      const result = runProfilerSweep();
+      expect(result.prunedCount).toBe(0);
+      expect(result.remainingRuns).toBe(2);
+
+      // Both directories still exist
+      expect(existsSync(join(runsDir, "run-1"))).toBe(true);
+      expect(existsSync(join(runsDir, "run-2"))).toBe(true);
+    });
+
+    test("prunes oldest completed runs when byte budget exceeded", () => {
+      // Set a very small byte budget
+      process.env.VELLUM_PROFILER_MAX_BYTES = "3000";
+      process.env.VELLUM_PROFILER_MAX_RUNS = "100";
+      process.env.VELLUM_PROFILER_MIN_FREE_MB = "0";
+
+      // Create runs with explicit timestamps for ordering
+      createRun("oldest", {
+        sizeBytes: 2000,
+        manifest: {
+          status: "completed",
+          createdAt: "2025-01-01T00:00:00Z",
+        },
+      });
+      createRun("middle", {
+        sizeBytes: 2000,
+        manifest: {
+          status: "completed",
+          createdAt: "2025-02-01T00:00:00Z",
+        },
+      });
+      createRun("newest", {
+        sizeBytes: 2000,
+        manifest: {
+          status: "completed",
+          createdAt: "2025-03-01T00:00:00Z",
+        },
+      });
+
+      const result = runProfilerSweep();
+
+      // Should prune until total bytes fit within 3000.
+      // Each run is ~2000 payload + manifest overhead. The sweep recomputes
+      // sizes so actual totals include the manifest file. At least 1 run
+      // should be pruned (the oldest).
+      expect(result.prunedCount).toBeGreaterThanOrEqual(1);
+      expect(result.freedBytes).toBeGreaterThan(0);
+
+      // The oldest should be gone
+      expect(existsSync(join(runsDir, "oldest"))).toBe(false);
+    });
+
+    test("prunes oldest completed runs when max-run-count exceeded", () => {
+      process.env.VELLUM_PROFILER_MAX_BYTES = "999999999";
+      process.env.VELLUM_PROFILER_MAX_RUNS = "2";
+      process.env.VELLUM_PROFILER_MIN_FREE_MB = "0";
+
+      createRun("run-a", {
+        sizeBytes: 100,
+        manifest: {
+          status: "completed",
+          createdAt: "2025-01-01T00:00:00Z",
+        },
+      });
+      createRun("run-b", {
+        sizeBytes: 100,
+        manifest: {
+          status: "completed",
+          createdAt: "2025-02-01T00:00:00Z",
+        },
+      });
+      createRun("run-c", {
+        sizeBytes: 100,
+        manifest: {
+          status: "completed",
+          createdAt: "2025-03-01T00:00:00Z",
+        },
+      });
+      createRun("run-d", {
+        sizeBytes: 100,
+        manifest: {
+          status: "completed",
+          createdAt: "2025-04-01T00:00:00Z",
+        },
+      });
+
+      const result = runProfilerSweep();
+
+      // 4 completed runs, max 2: should prune 2 oldest
+      expect(result.prunedCount).toBe(2);
+      expect(existsSync(join(runsDir, "run-a"))).toBe(false);
+      expect(existsSync(join(runsDir, "run-b"))).toBe(false);
+      expect(existsSync(join(runsDir, "run-c"))).toBe(true);
+      expect(existsSync(join(runsDir, "run-d"))).toBe(true);
+      expect(result.remainingRuns).toBe(2);
+    });
+
+    test("never deletes the active run", () => {
+      process.env.VELLUM_PROFILER_RUN_ID = "current";
+      process.env.VELLUM_PROFILER_MAX_BYTES = "500";
+      process.env.VELLUM_PROFILER_MAX_RUNS = "1";
+      process.env.VELLUM_PROFILER_MIN_FREE_MB = "0";
+
+      createRun("current", { sizeBytes: 2000 });
+      createRun("old-completed", {
+        sizeBytes: 2000,
+        manifest: {
+          status: "completed",
+          createdAt: "2025-01-01T00:00:00Z",
+        },
+      });
+
+      const result = runProfilerSweep();
+
+      // old-completed should be pruned, current should survive
+      expect(existsSync(join(runsDir, "current"))).toBe(true);
+      expect(existsSync(join(runsDir, "old-completed"))).toBe(false);
+      expect(result.prunedCount).toBe(1);
+    });
+
+    test("signals active-run-over-budget when active run exceeds byte budget", () => {
+      process.env.VELLUM_PROFILER_RUN_ID = "big-active";
+      process.env.VELLUM_PROFILER_MAX_BYTES = "500";
+      process.env.VELLUM_PROFILER_MAX_RUNS = "100";
+      process.env.VELLUM_PROFILER_MIN_FREE_MB = "0";
+
+      createRun("big-active", { sizeBytes: 10000 });
+
+      const result = runProfilerSweep();
+
+      expect(result.activeRunOverBudget).toBe(true);
+      // Active run must still exist
+      expect(existsSync(join(runsDir, "big-active"))).toBe(true);
+      expect(result.remainingRuns).toBe(1);
+    });
+
+    test("deletes single oversized completed run to recover space", () => {
+      process.env.VELLUM_PROFILER_MAX_BYTES = "100";
+      process.env.VELLUM_PROFILER_MAX_RUNS = "100";
+      process.env.VELLUM_PROFILER_MIN_FREE_MB = "0";
+
+      createRun("huge-completed", {
+        sizeBytes: 50000,
+        manifest: {
+          status: "completed",
+          createdAt: "2025-01-01T00:00:00Z",
+        },
+      });
+
+      const result = runProfilerSweep();
+
+      expect(result.prunedCount).toBe(1);
+      expect(result.freedBytes).toBeGreaterThanOrEqual(50000);
+      expect(existsSync(join(runsDir, "huge-completed"))).toBe(false);
+    });
+
+    test("creates profiler directories on first sweep if missing", () => {
+      // Remove everything
+      rmSync(join(testDir, "data", "profiler"), {
+        recursive: true,
+        force: true,
+      });
+
+      const result = runProfilerSweep();
+      expect(result.prunedCount).toBe(0);
+      expect(existsSync(runsDir)).toBe(true);
+    });
+
+    test("sweep is idempotent — repeated calls produce consistent state", () => {
+      process.env.VELLUM_PROFILER_MAX_BYTES = "999999";
+      process.env.VELLUM_PROFILER_MAX_RUNS = "10";
+      process.env.VELLUM_PROFILER_MIN_FREE_MB = "0";
+
+      createRun("stable-1", { sizeBytes: 1024 });
+      createRun("stable-2", { sizeBytes: 1024 });
+
+      const first = runProfilerSweep();
+      const second = runProfilerSweep();
+
+      expect(first.prunedCount).toBe(0);
+      expect(second.prunedCount).toBe(0);
+      expect(first.remainingRuns).toBe(second.remainingRuns);
+    });
+
+    test("active run is not counted against max completed runs", () => {
+      process.env.VELLUM_PROFILER_RUN_ID = "live";
+      process.env.VELLUM_PROFILER_MAX_BYTES = "999999";
+      process.env.VELLUM_PROFILER_MAX_RUNS = "2";
+      process.env.VELLUM_PROFILER_MIN_FREE_MB = "0";
+
+      createRun("live", { sizeBytes: 100 });
+      createRun("done-1", {
+        sizeBytes: 100,
+        manifest: {
+          status: "completed",
+          createdAt: "2025-01-01T00:00:00Z",
+        },
+      });
+      createRun("done-2", {
+        sizeBytes: 100,
+        manifest: {
+          status: "completed",
+          createdAt: "2025-02-01T00:00:00Z",
+        },
+      });
+
+      const result = runProfilerSweep();
+
+      // 2 completed runs = max, so nothing should be pruned
+      expect(result.prunedCount).toBe(0);
+      // Active + 2 completed = 3 remaining
+      expect(result.remainingRuns).toBe(3);
+      expect(existsSync(join(runsDir, "live"))).toBe(true);
+      expect(existsSync(join(runsDir, "done-1"))).toBe(true);
+      expect(existsSync(join(runsDir, "done-2"))).toBe(true);
+    });
+  });
+});

package/src/__tests__/proxy-approval-callback.test.ts

@@ -189,7 +189,7 @@ describe("createProxyApprovalCallback", () => {
     expect(prompterSendToClient).not.toHaveBeenCalled();
   });
 
-  test("high-risk with plain allow rule (no allowHighRisk) falls through to prompt", async () => {
+  test("ask_missing_credential with allow rule auto-allows (medium risk)", async () => {
     findHighestPriorityRuleMock.mockReturnValue({
       id: "rule-hr-1",
       tool: "network_request",
@@ -198,53 +198,17 @@ describe("createProxyApprovalCallback", () => {
       decision: "allow" as const,
       priority: 100,
       createdAt: Date.now(),
-      // No allowHighRisk — should NOT auto-allow for high-risk decisions
     });
 
     const ctx = makeContext();
     const prompterSendToClient = mock(() => {});
     const prompter = new PermissionPrompter(prompterSendToClient);
 
-    const originalPrompt = prompter.prompt.bind(prompter);
-    prompter.prompt = async (...args) => {
-      const p = originalPrompt(...args);
-      await new Promise((r) => setTimeout(r, 10));
-      const call = (prompterSendToClient.mock.calls as unknown[][])[0];
-      const msg = call[0] as { requestId: string };
-      prompter.resolveConfirmation(msg.requestId, "allow");
-      return p;
-    };
-
     const callback = createProxyApprovalCallback(prompter, ctx);
-    // ask_missing_credential is high risk
     const result = await callback(makeAskMissingCredentialRequest());
 
     expect(result).toBe(true);
-    // Prompter SHOULD have been called — plain allow rule doesn't auto-allow high-risk
-    expect(prompterSendToClient).toHaveBeenCalled();
-  });
-
-  test("high-risk with allowHighRisk allow rule auto-allows without prompting", async () => {
-    findHighestPriorityRuleMock.mockReturnValue({
-      id: "rule-hr-2",
-      tool: "network_request",
-      pattern: "network_request:https://api.fal.ai:443/*",
-      scope: "/tmp/test-project",
-      decision: "allow" as const,
-      priority: 100,
-      createdAt: Date.now(),
-      allowHighRisk: true,
-    });
-
-    const ctx = makeContext();
-    const prompterSendToClient = mock(() => {});
-    const prompter = new PermissionPrompter(prompterSendToClient);
-
-    const callback = createProxyApprovalCallback(prompter, ctx);
-    const result = await callback(makeAskMissingCredentialRequest());
-
-    expect(result).toBe(true);
-    // Prompter should NOT have been called — allowHighRisk rule auto-allows
+    // Plain allow rule auto-allows medium-risk requests
     expect(prompterSendToClient).not.toHaveBeenCalled();
   });
 
@@ -436,7 +400,7 @@ describe("createProxyApprovalCallback", () => {
     await callback(makeAskUnauthenticatedRequest());
   });
 
-  test("uses high risk level for ask_missing_credential decisions", async () => {
+  test("uses medium risk level for ask_missing_credential decisions", async () => {
     const ctx = makeContext();
     const prompterSendToClient = mock(() => {});
     const prompter = new PermissionPrompter(prompterSendToClient);
@@ -447,8 +411,7 @@ describe("createProxyApprovalCallback", () => {
       await new Promise((r) => setTimeout(r, 10));
       const call = (prompterSendToClient.mock.calls as unknown[][])[0];
       const msg = call[0] as { requestId: string; riskLevel: string };
-      // Missing credential prompts are high risk — the target wants auth
-      expect(msg.riskLevel).toBe("high");
+      expect(msg.riskLevel).toBe("medium");
       prompter.resolveConfirmation(msg.requestId, "allow");
       return p;
     };
@@ -523,40 +486,6 @@ describe("createProxyApprovalCallback", () => {
   // in contrast to the proxied bash activation path which CANNOT (tested
   // in tool-executor.test.ts).
 
-  test("always_allow_high_risk persists rule with allowHighRisk flag", async () => {
-    const ctx = makeContext();
-    const prompterSendToClient = mock(() => {});
-    const prompter = new PermissionPrompter(prompterSendToClient);
-
-    const originalPrompt = prompter.prompt.bind(prompter);
-    prompter.prompt = async (...args) => {
-      const p = originalPrompt(...args);
-      await new Promise((r) => setTimeout(r, 10));
-      const call = (prompterSendToClient.mock.calls as unknown[][])[0];
-      const msg = call[0] as { requestId: string };
-      prompter.resolveConfirmation(
-        msg.requestId,
-        "always_allow_high_risk",
-        "network_request:https://api.fal.ai:443/*",
-        "/tmp/test-project",
-      );
-      return p;
-    };
-
-    const callback = createProxyApprovalCallback(prompter, ctx);
-    const result = await callback(makeAskMissingCredentialRequest());
-
-    expect(result).toBe(true);
-    expect(addRuleMock).toHaveBeenCalledWith(
-      "network_request",
-      "network_request:https://api.fal.ai:443/*",
-      "/tmp/test-project",
-      "allow",
-      100,
-      { allowHighRisk: true },
-    );
-  });
-
   test("one-time allow does NOT persist any rule", async () => {
     const ctx = makeContext();
     const prompterSendToClient = mock(() => {});

package/src/__tests__/registry.test.ts

@@ -109,7 +109,7 @@ describe("tool registry dynamic-tools tools", () => {
 
 describe("tool manifest", () => {
   test("eager module tool names list contains expected count", () => {
-    expect(eagerModuleToolNames.length).toBe(9);
+    expect(eagerModuleToolNames.length).toBe(11);
   });
 
   test("explicit tools list includes memory and credential tools", () => {

package/src/__tests__/require-fresh-approval.test.ts

@@ -416,7 +416,6 @@ describe("requireFreshApproval: persistent decisions disabled", () => {
         _allowlistOptions: unknown[],
         _scopeOptions: unknown[],
         _previewDiff: unknown,
-        _sandboxed: boolean | undefined,
         _conversationId: string,
         _executionTarget: string,
         persistentDecisionsAllowed: boolean,
@@ -467,7 +466,6 @@ describe("requireFreshApproval: persistent decisions disabled", () => {
         _allowlistOptions: unknown[],
         _scopeOptions: unknown[],
         _previewDiff: unknown,
-        _sandboxed: boolean | undefined,
         _conversationId: string,
         _executionTarget: string,
         persistentDecisionsAllowed: boolean,

package/src/__tests__/sandbox-diagnostics.test.ts

@@ -10,22 +10,6 @@ mock.module("node:child_process", () => ({
   execSync: execSyncMock,
 }));
 
-// Mock config loader — return a config with sandbox settings
-let mockSandboxConfig: {
-  enabled: boolean;
-} = {
-  enabled: true,
-};
-
-mock.module("../config/loader.js", () => ({
-  getConfig: () => ({
-    ui: {},
-
-    sandbox: mockSandboxConfig,
-  }),
-  loadRawConfig: () => ({}),
-}));
-
 mock.module("../util/logger.js", () => ({
   getLogger: () => ({
     error: () => {},
@@ -49,9 +33,6 @@ function setPlatform(platform: string): void {
 
 beforeEach(() => {
   execSyncMock.mockReset();
-  mockSandboxConfig = {
-    enabled: true,
-  };
   // Default: all commands succeed.
   execSyncMock.mockImplementation(() => undefined);
 });
@@ -62,26 +43,14 @@ afterEach(() => {
 });
 
 describe("runSandboxDiagnostics — config reporting", () => {
-  test("reports sandbox enabled state", () => {
-    const result = runSandboxDiagnostics();
-    expect(result.config.enabled).toBe(true);
-  });
-
-  test("reports sandbox disabled state", () => {
-    mockSandboxConfig.enabled = false;
+  test("reports sandbox disabled (sandbox removed)", () => {
     const result = runSandboxDiagnostics();
     expect(result.config.enabled).toBe(false);
   });
 });
 
 describe("runSandboxDiagnostics — active backend reason", () => {
-  test("explains native backend selection", () => {
-    const result = runSandboxDiagnostics();
-    expect(result.activeBackendReason).toContain("Native backend");
-  });
-
   test("explains when sandbox is disabled", () => {
-    mockSandboxConfig.enabled = false;
     const result = runSandboxDiagnostics();
     expect(result.activeBackendReason).toContain("disabled");
   });

package/src/__tests__/sandbox-host-parity.test.ts

@@ -705,10 +705,11 @@ describe("Terminal output format: formatShellOutput shared by sandbox and host",
     const longOutput = "x".repeat(MAX_OUTPUT_LENGTH + 100);
     const result = formatShellOutput(longOutput, "", 0, false, 120);
 
-    expect(result.content.length).toBe(
-      MAX_OUTPUT_LENGTH + 1 + '<output_truncated limit="50K" />'.length,
-    );
-    expect(result.content).toContain("<output_truncated");
+    expect(result.content).toContain('limit="20K"');
+    expect(result.content).toContain('file="');
+    // The <output_truncated tag starts right after MAX_OUTPUT_LENGTH chars + 1 newline
+    const tagStart = result.content.indexOf("<output_truncated");
+    expect(tagStart).toBe(MAX_OUTPUT_LENGTH + 1);
   });
 
   test("timed-out command appends timeout tag and sets isError", () => {