diff --git a/src/artifacts/NtfsMftArtifactReader.test.ts b/src/artifacts/NtfsMftArtifactReader.test.ts
index 5dcc2891cb..944af32087 100644
--- a/src/artifacts/NtfsMftArtifactReader.test.ts
+++ b/src/artifacts/NtfsMftArtifactReader.test.ts
@@ -9,6 +9,7 @@ import { NtfsMftArtifactReader } from "./NtfsMftArtifactReader.js";
 
 import { MFT_DEFAULT_RECORD_BYTES } from "../domain/windows/ntfsMft.js";
 import { createTestTempDirectory } from "../../tests/fixtures/temporaryDirectory.js";
+import { linkedMftFixture } from "../../tests/fixtures/ntfsMftExtension.js";
 import { scanCanonicalArtifactInventory } from "../../tests/fixtures/artifactInventory.js";
 
 /** A 1 KiB FILE record whose update sequence array covers both sectors. */
@@ -77,7 +78,7 @@ it("reports a $MFT stream's record headers as the root occurrence", async () =>
     /attributes, filenames.*not decoded|no path, file size/iu,
   );
   expect(limitations).toContain(
-    "Nonresident payloads and extension records are not assembled",
+    "Nonresident list values and file contents are not recovered",
   );
   // A record stream declares no members, so the root is its whole inventory.
   expect(inventory.occurrences.map((row) => row.logical_path)).toEqual(["."]);
@@ -240,3 +241,35 @@ it("revalidates the source before reusing a completed header inventory", async (
     await reader.close();
   }
 });
+
+it("delivers acquired extension metadata through the canonical inventory schema", async () => {
+  const inventory = await scanCanonicalArtifactInventory(
+    await writeMft("linked", linkedMftFixture()),
+  );
+  expect(inventory.occurrences[0]).toMatchObject({
+    artifact_format: "ntfs-mft",
+    format_metadata: {
+      record_count: 2,
+      extension_record_count: 1,
+      file_name_count: 2,
+      record_headers: [
+        {
+          index: 0,
+          fileNames: [{ name: "local" }],
+          attributeListResolution: {
+            coverage: "complete",
+            references: [
+              { result: { kind: "resolved", attributeFileOffset: 1080 } },
+            ],
+            linkedFileNames: [{ name: "linked", fileOffset: 1104 }],
+          },
+        },
+        {
+          index: 1,
+          baseRecordReference: { recordNumber: 0, sequenceNumber: 2 },
+          fileNames: [{ name: "linked" }],
+        },
+      ],
+    },
+  });
+});
diff --git a/src/artifacts/artifactFormats.ts b/src/artifacts/artifactFormats.ts
index 78e815bdeb..d9b3fdcc2a 100644
--- a/src/artifacts/artifactFormats.ts
+++ b/src/artifacts/artifactFormats.ts
@@ -470,7 +470,7 @@ const NON_ZIP_FORMATS = {
     ...plain,
     retainClassifiedSource: true,
     readerLimitation:
-      "MFT inventory reports whole-stream counts and up to 32 FILE record details with fixup coverage, supported filenames, timestamps, declared sizes and attribute metadata; omitted details are counted. Nonresident payloads and extension records are not assembled, full paths are not reconstructed, and MFT inventory alone does not establish change history. Use inspect_ntfs_journal with caller-paired acquisitions for bounded journal events.",
+      "MFT inventory reports whole-stream counts and up to 32 physical FILE record details, including fixup coverage, supported names, timestamps, declared sizes and attributes. Bounded attribute-list links expose verified metadata from acquired extension records, with explicit reference outcomes and omissions. Nonresident list values and file contents are not recovered; full paths and complete change history are not established. Use inspect_ntfs_journal for caller-paired journal events.",
     createReader: (path, options) =>
       new NtfsMftArtifactReader(path, sourceOf(options)),
   }),
diff --git a/tests/boundary/mcp/ntfsJournal.test.ts b/tests/boundary/mcp/ntfsJournal.test.ts
index 6a3e66b710..d7b78280e4 100644
--- a/tests/boundary/mcp/ntfsJournal.test.ts
+++ b/tests/boundary/mcp/ntfsJournal.test.ts
@@ -2,6 +2,8 @@ import { writeFile } from "node:fs/promises";
 import { join } from "node:path";
 
 import { expect, vi } from "vitest";
+import { z } from "zod";
+import { linkedMftFixture } from "../../fixtures/ntfsMftExtension.js";
 import * as stableArtifact from "../../../src/artifacts/readStableArtifact.js";
 import * as mftDecoder from "../../../src/domain/windows/ntfsMft.js";
 
@@ -157,7 +159,8 @@ mcpTest(
     if (!config.ok) throw config.error;
     const session = createBinarySession(config.value, silentLogger, {});
     onTestFinished(async () => {
-      await session.close();
+      const closed = await session.close();
+      if (!closed.ok) throw closed.error;
     });
     const client = await mcp.connect(
       createServer({ kind: "session", session }),
@@ -198,14 +201,9 @@ mcpTest(
     });
     expect(refused.isError).toBe(true);
     const { error } = parseMcpToolError(refused);
-    const reference = error.details?.partial_observation;
-    if (
-      typeof reference !== "object" ||
-      reference === null ||
-      !("evidence_id" in reference) ||
-      typeof reference.evidence_id !== "string"
-    )
-      throw new Error("Missing retained MFT observation");
+    const reference = z
+      .object({ evidence_id: z.string() })
+      .parse(error.details?.partial_observation);
     const partial = parseEvidence(session.evidenceById(reference.evidence_id));
     expect(partial.normalized_result).toMatchObject({
       kind: "ntfs-journal-partial",
@@ -219,3 +217,102 @@ mcpTest(
     });
   },
 );
+
+mcpTest(
+  "CLI/MCP retain extension facts through independent detail pages and later journal failure",
+  async ({ mcp, onTestFinished }) => {
+    const directory = await createTestTempDirectory("rea-ntfs-linked-");
+    const mftPath = join(directory, "mft"),
+      journalPath = join(directory, "journal");
+    const journal = Buffer.alloc(64);
+    journal.writeUInt32LE(64);
+    journal.writeUInt16LE(2, 4);
+    journal.writeBigUInt64LE(2n << 48n, 8);
+    journal.writeUInt16LE(60, 58);
+    await writeFile(mftPath, linkedMftFixture());
+    await writeFile(journalPath, journal);
+    const cli = await serveCli([
+      "inspect-ntfs-journal",
+      mftPath,
+      journalPath,
+      "--mftRecordLimit=1",
+      "--json",
+    ]);
+    expect(cli.exitCode, cli.stdout).toBe(0);
+    const expected = ntfsJournalResultSchema.parse(
+      parseEvidence(JSON.parse(cli.stdout)).normalized_result,
+    );
+    expect(expected.mftMetadata).toMatchObject({
+      record_count: 2,
+      file_name_count: 2,
+      next_record_header_index: 1,
+      record_headers: [
+        {
+          fileNames: [{ name: "local" }],
+          attributeListResolution: {
+            coverage: "complete",
+            linkedFileNames: [{ name: "linked", fileOffset: 1104 }],
+          },
+        },
+      ],
+    });
+    const config = parseConfig({});
+    if (!config.ok) throw config.error;
+    const session = createBinarySession(config.value, silentLogger, {});
+    onTestFinished(async () => {
+      const closed = await session.close();
+      if (!closed.ok) throw closed.error;
+    });
+    const client = await mcp.connect(
+      createServer({ kind: "session", session }),
+    );
+    const reads = vi.spyOn(stableArtifact, "readStableArtifact"),
+      parses = vi.spyOn(mftDecoder, "parseMftIndex");
+    onTestFinished(() => {
+      reads.mockRestore();
+      parses.mockRestore();
+    });
+    const input = {
+      mft_path: mftPath,
+      journal_path: journalPath,
+      mft_record_limit: 1,
+    };
+    const reply = await client.callTool({
+      name: "inspect_ntfs_journal",
+      arguments: input,
+    });
+    expect(reply.isError).not.toBe(true);
+    expect(parseEvidence(reply.structuredContent).normalized_result).toEqual(
+      expected,
+    );
+    const next = await client.callTool({
+      name: "inspect_ntfs_journal",
+      arguments: { ...input, mft_record_start: 1, start_file_offset: 64 },
+    });
+    expect(next.isError).not.toBe(true);
+    expect(
+      parseEvidence(next.structuredContent).normalized_result,
+    ).toMatchObject({
+      rows: [],
+      mftMetadata: {
+        record_headers: [{ index: 1, fileNames: [{ name: "linked" }] }],
+        next_record_header_index: null,
+      },
+    });
+    expect(reads).toHaveBeenCalledTimes(2);
+    expect(parses).toHaveBeenCalledTimes(1);
+    const refused = await client.callTool({
+      name: "inspect_ntfs_journal",
+      arguments: { ...input, journal_path: join(directory, "missing") },
+    });
+    expect(refused.isError).toBe(true);
+    const { error } = parseMcpToolError(refused);
+    const reference = z
+      .object({ evidence_id: z.string() })
+      .parse(error.details?.partial_observation);
+    expect(
+      parseEvidence(session.evidenceById(reference.evidence_id))
+        .normalized_result,
+    ).toMatchObject({ mft: { metadata: expected.mftMetadata } });
+  },
+);
diff --git a/tests/fixtures/ntfsMftExtension.ts b/tests/fixtures/ntfsMftExtension.ts
index c1f133958a..be0e033ea0 100644
--- a/tests/fixtures/ntfsMftExtension.ts
+++ b/tests/fixtures/ntfsMftExtension.ts
@@ -74,6 +74,13 @@ export const mftFileRecord = (
   record.writeUInt16LE(1, 22);
   record.writeUInt32LE(recordBytes, 28);
   record.writeBigUInt64LE(baseReference, 32);
+  record.writeUInt16LE(
+    attributes.reduce(
+      (next, attribute) => Math.max(next, attribute.readUInt16LE(14) + 1),
+      0,
+    ),
+    40,
+  );
   record.writeUInt32LE(index, 44);
   let cursor = first;
   for (const attribute of attributes) {