test(mcp): покрыть error-ветки drawio-тулов + escaping round-trip; удалить мёртвый freshBlockId (ревью #434)

- drawio_get: битый ref -> 'no node found'; drawio-нода без src -> 'has no src';
  drawio_update: узел не drawio -> чистая ошибка + ноль upload/mutate;
  drawio_create: anchor не найден -> ошибка 'unreferenced orphan' + attachment
  залит и назван в сообщении, drawio-узел не записан.
- escaping round-trip: title с < > " & переживает encode/build/decode байт-в-байт,
  внешний content="..." остаётся well-formed.
- удалён неиспользуемый freshBlockId() (0 call-site после фикса #index-хендла).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

apps/server/src/core/ai-chat/tools/ai-chat-tools.service.ts

@@ -17,7 +17,7 @@ import {
   resolveCurrentPageResult,
   type SelectionContext,
 } from './current-page.util';
-import { parseNodeArg } from '@docmost/prosemirror-markdown';
+import { parseNodeArg } from './parse-node-arg';
 import { modelFriendlyInput } from './model-friendly-input';
 import { SandboxStore } from '../../../integrations/sandbox/sandbox.store';
 import {

apps/server/src/core/ai-chat/tools/parse-node-arg.spec.ts

@@ -1,10 +1,10 @@
-import { parseNodeArg } from '@docmost/prosemirror-markdown';
+import { parseNodeArg } from './parse-node-arg';
 
 /**
- * Unit tests for the shared `parseNodeArg` helper (#414: now the single copy in
- * `@docmost/prosemirror-markdown`, imported by both the server tool adapters and
- * `@docmost/mcp`). Used by the patchNode / insertNode / updatePageJson adapters.
- * Behavior: object passthrough, valid-string parse, invalid-string throw.
+ * Unit tests for the in-app `parseNodeArg` helper. It mirrors the standalone
+ * MCP helper (packages/mcp/src/lib/parse-node-arg.ts) and is used by the
+ * patchNode / insertNode / updatePageJson tool adapters. Behavior must be
+ * byte-identical: object passthrough, valid-string parse, invalid-string throw.
  */
 describe('parseNodeArg', () => {
   it('passes an object through unchanged', () => {

apps/server/src/core/ai-chat/tools/parse-node-arg.ts

@@ -0,0 +1,26 @@
+// The model sometimes serializes a ProseMirror node arg as a JSON string
+// instead of an object. Normalize: parse a string to an object (throwing on
+// invalid JSON), pass an object through unchanged. Shared by patchNode /
+// insertNode (and the analogous updatePageJson content parsing).
+//
+// This is behaviorally identical to `packages/mcp/src/lib/parse-node-arg.ts`
+// (the function logic, default/explicit throw messages and branch order match;
+// only comments and quote style differ). We cannot import that helper here:
+// `@docmost/mcp` is ESM-only and this server
+// compiles with module:commonjs, so it is loaded at runtime via the
+// `new Function('import()')` trick (see docmost-client.loader.ts). Sharing
+// runtime code across that ESM/CJS boundary by a normal import is impossible,
+// hence the mirrored copy.
+export function parseNodeArg(
+  node: unknown,
+  errMsg = 'node was a string but not valid JSON',
+): unknown {
+  if (typeof node === 'string') {
+    try {
+      return JSON.parse(node);
+    } catch {
+      throw new Error(errMsg);
+    }
+  }
+  return node;
+}

packages/mcp/package.json

@@ -52,6 +52,7 @@
     "form-data": "^4.0.0",
     "jsdom": "^27.4.0",
     "marked": "^17.0.1",
+    "pako": "^2.0.3",
     "re2": "^1.21.0",
     "ws": "^8.19.0",
     "y-prosemirror": "1.3.7",

packages/mcp/src/client.ts

@@ -46,9 +46,18 @@ import {
   insertTableRow,
   deleteTableRow,
   updateTableCell,
-} from "@docmost/prosemirror-markdown";
+} from "./lib/node-ops.js";
 import { searchInDoc, SearchOptions } from "./lib/page-search.js";
 import { withPageLock } from "./lib/page-lock.js";
+import {
+  prepareModel,
+  decodeDrawioSvg,
+  buildDrawioSvg,
+  mxHash,
+  normalizeXml,
+  countUserCells,
+} from "./lib/drawio-xml.js";
+import { renderDiagramShapes } from "./lib/drawio-preview.js";
 import {
   applyTextEdits,
   TextEdit,
@@ -3432,6 +3441,453 @@ export class DocmostClient {
     });
   }
 
+  // --- draw.io diagrams (issue #423) ---
+
+  /**
+   * Upload a ready-made byte buffer as a page attachment via the same
+   * multipart /files/upload endpoint uploadImage uses. Split out as its own
+   * (overridable) seam so drawio_create/update can upload the generated
+   * `.drawio.svg` without going through the URL-fetch path, and so tests can
+   * stub the network. Mirrors uploadImage's fresh-FormData + one-shot 401/403
+   * re-auth handling (a FormData body is single-use, so it must be rebuilt per
+   * attempt).
+   */
+  protected async uploadAttachmentBuffer(
+    pageId: string,
+    buffer: Buffer,
+    fileName: string,
+    mime: string,
+  ): Promise<{ id: string; fileName: string; fileSize: number }> {
+    await this.ensureAuthenticated();
+    const buildForm = () => {
+      const form = new FormData();
+      form.append("pageId", pageId);
+      form.append("file", buffer, { filename: fileName, contentType: mime });
+      return form;
+    };
+    const uploadUrl = `${this.apiUrl}/files/upload`;
+    let response;
+    try {
+      const form = buildForm();
+      response = await axios.post(uploadUrl, form, {
+        headers: {
+          ...form.getHeaders(),
+          Authorization: this.client.defaults.headers.common["Authorization"],
+        },
+        timeout: 60000,
+      });
+    } catch (error) {
+      if (
+        axios.isAxiosError(error) &&
+        (error.response?.status === 401 || error.response?.status === 403)
+      ) {
+        await this.login();
+        const form2 = buildForm();
+        response = await axios.post(uploadUrl, form2, {
+          headers: {
+            ...form2.getHeaders(),
+            Authorization: this.client.defaults.headers.common["Authorization"],
+          },
+          timeout: 60000,
+        });
+      } else if (axios.isAxiosError(error)) {
+        if (process.env.DEBUG) {
+          console.error(
+            "Attachment upload failed; response body:",
+            JSON.stringify(error.response?.data),
+          );
+        }
+        throw new Error(
+          `Attachment upload failed: ${error.response?.status} ${error.response?.statusText}`,
+        );
+      } else {
+        throw error;
+      }
+    }
+    const att = response.data?.data ?? response.data;
+    if (!att?.id || !att?.fileName) {
+      throw new Error(
+        "Unexpected /files/upload response: " + JSON.stringify(response.data),
+      );
+    }
+    return {
+      id: att.id,
+      fileName: att.fileName,
+      fileSize: att.fileSize ?? buffer.length,
+    };
+  }
+
+  /**
+   * Fetch a stored `.drawio.svg` attachment as text. Overridable seam over
+   * fetchInternalFile (the authed loopback fetch, which also rejects any
+   * traversal/SSRF src) so drawio_get/update can read the current diagram and
+   * tests can stub the bytes.
+   */
+  protected async fetchAttachmentText(src: string): Promise<string> {
+    const { buffer } = await this.fetchInternalFile(src);
+    return buffer.toString("utf-8");
+  }
+
+  /**
+   * Resolve a drawio node on a page by `attrs.id` or `#<index>` and return the
+   * node plus its ref. Throws a clear error if the ref does not resolve to a
+   * drawio node.
+   */
+  private async resolveDrawioNode(
+    pageId: string,
+    node: string,
+  ): Promise<{ node: any; ref: string }> {
+    const data = await this.getPageRaw(pageId);
+    const hit = getNodeByRef(
+      data.content ?? { type: "doc", content: [] },
+      node,
+    );
+    if (!hit) {
+      throw new Error(
+        `drawio: no node found for "${node}" on page ${pageId} (use the drawio node's attrs.id or "#<index>" from get_outline)`,
+      );
+    }
+    if (hit.type !== "drawio") {
+      throw new Error(
+        `drawio: node "${node}" on page ${pageId} is a ${hit.type}, not a drawio diagram`,
+      );
+    }
+    return { node: hit.node, ref: node };
+  }
+
+  /**
+   * Read a drawio diagram as mxGraph XML (default) or as the raw `.drawio.svg`.
+   * Runs the decode chain (base64/entity content= → drawio file → nested XML or
+   * pako-inflated compressed <diagram>). The returned `hash` is the
+   * optimistic-lock key for drawio_update.
+   */
+  async drawioGet(
+    pageId: string,
+    node: string,
+    format: "xml" | "svg" = "xml",
+  ): Promise<{
+    pageId: string;
+    nodeId: string;
+    format: "xml" | "svg";
+    content: string;
+    meta: {
+      attachmentId: string | null;
+      title: string | null;
+      width: number | null;
+      height: number | null;
+      cellCount: number;
+      hash: string;
+    };
+  }> {
+    await this.ensureAuthenticated();
+    const { node: drawio } = await this.resolveDrawioNode(pageId, node);
+    const attrs = drawio.attrs || {};
+    const src = attrs.src;
+    if (!src) {
+      throw new Error(
+        `drawio: node "${node}" on page ${pageId} has no src to read`,
+      );
+    }
+    const svg = await this.fetchAttachmentText(src);
+    const modelXml = decodeDrawioSvg(svg);
+    const meta = {
+      attachmentId: attrs.attachmentId ?? null,
+      title: attrs.title ?? null,
+      width: attrs.width != null ? Number(attrs.width) : null,
+      height: attrs.height != null ? Number(attrs.height) : null,
+      cellCount: countUserCells(modelXml),
+      hash: mxHash(modelXml),
+    };
+    return {
+      pageId,
+      nodeId: attrs.id ?? node,
+      format,
+      content: format === "svg" ? svg : normalizeXml(modelXml),
+      meta,
+    };
+  }
+
+  /**
+   * Create a drawio diagram from mxGraph XML: lint → schematic SVG preview
+   * (pure TS) → build the `.drawio.svg` (createDrawioSvg contract) → create the
+   * attachment → insert a `drawio` node before/after an anchor or appended.
+   * `xml` is a bare `<mxGraphModel>` or a list of `<mxCell>` (the server wraps
+   * it and adds the id=0/id=1 sentinels).
+   */
+  async drawioCreate(
+    pageId: string,
+    where: {
+      position: "before" | "after" | "append";
+      anchorNodeId?: string;
+      anchorText?: string;
+    },
+    xml: string,
+    title?: string,
+  ): Promise<{
+    success: boolean;
+    nodeId: string;
+    attachmentId: string;
+    warnings: string[];
+    verify?: any;
+  }> {
+    await this.ensureAuthenticated();
+    if (
+      !where ||
+      (where.position !== "before" &&
+        where.position !== "after" &&
+        where.position !== "append")
+    ) {
+      throw new Error(
+        'drawio_create: `where.position` must be one of "before", "after", "append"',
+      );
+    }
+    if (where.position === "before" || where.position === "after") {
+      const hasId =
+        typeof where.anchorNodeId === "string" && where.anchorNodeId.length > 0;
+      const hasText =
+        typeof where.anchorText === "string" && where.anchorText.length > 0;
+      if (hasId === hasText) {
+        throw new Error(
+          `drawio_create: position "${where.position}" requires exactly one of anchorNodeId or anchorText`,
+        );
+      }
+    }
+
+    // Pre-write pipeline (throws a structured DrawioLintError on any violation).
+    const prepared = prepareModel(xml);
+    const inner = renderDiagramShapes(prepared.cells, prepared.bbox);
+    const diagramTitle = title || "Page-1";
+    const svg = buildDrawioSvg(prepared.modelXml, inner, prepared.bbox, diagramTitle);
+
+    const att = await this.uploadAttachmentBuffer(
+      pageId,
+      Buffer.from(svg, "utf-8"),
+      "diagram.drawio.svg",
+      "image/svg+xml",
+    );
+
+    // NOTE: no `id` attribute is set here. The vendored `drawio` node schema
+    // (diagramAttributes) declares no `id`, so any block id would be silently
+    // dropped by PMNode.fromJSON on save and the returned handle would fail to
+    // resolve. The addressable handle is the node's "#<index>" (like image/table
+    // nodes), computed after the insert below.
+    const drawioNode: any = {
+      type: "drawio",
+      attrs: {
+        src: `/api/files/${att.id}/${att.fileName}`,
+        attachmentId: att.id,
+        width: prepared.bbox.width,
+        height: prepared.bbox.height,
+        align: "center",
+      },
+    };
+    if (title) drawioNode.attrs.title = title;
+    // Reuse the existing URL trust boundary (rejects unsafe src schemes).
+    this.validateDocUrls(drawioNode);
+
+    const collabToken = await this.getCollabTokenWithReauth();
+    const pageUuid = await this.resolvePageId(pageId);
+
+    let inserted = false;
+    let insertedIndex = -1;
+    const mutation = await this.mutatePage(
+      pageUuid,
+      collabToken,
+      this.apiUrl,
+      (liveDoc) => {
+        inserted = false;
+        insertedIndex = -1;
+        const { doc: nd, inserted: ins } = insertNodeRelative(
+          liveDoc,
+          drawioNode,
+          where,
+        );
+        inserted = ins;
+        if (!inserted) return null; // anchor not found -> skip the write
+        // Locate the freshly-inserted node to derive its "#<index>" handle. The
+        // just-uploaded attachmentId is unique, so it identifies our node.
+        if (Array.isArray(nd.content)) {
+          insertedIndex = nd.content.findIndex(
+            (b: any) =>
+              b &&
+              b.type === "drawio" &&
+              b.attrs &&
+              b.attrs.attachmentId === att.id,
+          );
+        }
+        return nd;
+      },
+    );
+
+    if (!inserted) {
+      const anchorDesc = where.anchorNodeId
+        ? `anchorNodeId "${where.anchorNodeId}"`
+        : `anchorText "${where.anchorText}"`;
+      throw new Error(
+        `drawio_create: anchor not found (${anchorDesc}) on page ${pageId}. The diagram attachment ${att.id} is now an unreferenced orphan.`,
+      );
+    }
+
+    if (insertedIndex < 0) {
+      // The node was inserted nested (e.g. inside a callout/table cell via an
+      // anchor), where "#<index>" — which addresses only top-level blocks —
+      // cannot reference it. drawio nodes carry no persisted id, so there is no
+      // stable handle for a nested diagram.
+      throw new Error(
+        `drawio_create: the diagram was inserted on page ${pageId} but not as a ` +
+          `top-level block, so it has no addressable "#<index>" handle. Anchor ` +
+          `on a top-level block (or append) so the diagram can be re-read.`,
+      );
+    }
+
+    // The returned handle is POSITIONAL ("#<index>"): valid for the immediate
+    // create -> get/update flow, but re-resolve via get_outline if the document
+    // structure changes (blocks added/removed before it shift the index).
+    const nodeId = `#${insertedIndex}`;
+
+    return {
+      success: true,
+      nodeId,
+      attachmentId: att.id,
+      warnings: prepared.warnings,
+      verify: mutation.verify,
+    };
+  }
+
+  /**
+   * Full-replacement update of a drawio diagram. `baseHash` is MANDATORY: it is
+   * compared against the hash of the diagram's CURRENT XML (from drawio_get);
+   * any mismatch means a human or another agent edited the diagram after the
+   * read, so the write is refused with a conflict error. On success the new
+   * `.drawio.svg` is uploaded as a FRESH attachment (in-place byte overwrite is
+   * avoided — some Docmost versions corrupt an attachment on overwrite, exactly
+   * as replaceImage documents) and the node is repointed with new dimensions.
+   */
+  async drawioUpdate(
+    pageId: string,
+    node: string,
+    xml: string,
+    baseHash: string,
+  ): Promise<{
+    success: boolean;
+    nodeId: string;
+    attachmentId: string;
+    warnings: string[];
+    verify?: any;
+  }> {
+    await this.ensureAuthenticated();
+    if (typeof baseHash !== "string" || baseHash.length === 0) {
+      throw new Error(
+        "drawio_update: baseHash is mandatory — read the diagram with drawio_get first and pass back its meta.hash",
+      );
+    }
+
+    // Resolve the node and read the CURRENT diagram to enforce the optimistic
+    // lock before doing any write or upload.
+    const { node: drawio, ref } = await this.resolveDrawioNode(pageId, node);
+    const oldAttrs = drawio.attrs || {};
+    const oldSrc = oldAttrs.src;
+    // The returned handle is the caller-supplied reference. drawio nodes carry
+    // no persisted id, so `ref` (an "#<index>" or a rare legacy attrs.id) is the
+    // honest identifier to hand back.
+    const nodeId = oldAttrs.id ?? ref;
+    if (!oldSrc) {
+      throw new Error(
+        `drawio_update: node "${node}" on page ${pageId} has no src to compare against`,
+      );
+    }
+    const currentSvg = await this.fetchAttachmentText(oldSrc);
+    const currentHash = mxHash(decodeDrawioSvg(currentSvg));
+    if (currentHash !== baseHash) {
+      throw new Error(
+        `drawio_update: conflict — the diagram changed since it was read ` +
+          `(baseHash ${baseHash} != current ${currentHash}). Re-read it with drawio_get and retry.`,
+      );
+    }
+
+    // Pipeline for the new content (throws a structured DrawioLintError).
+    const prepared = prepareModel(xml);
+    const inner = renderDiagramShapes(prepared.cells, prepared.bbox);
+    const diagramTitle = oldAttrs.title || "Page-1";
+    const svg = buildDrawioSvg(prepared.modelXml, inner, prepared.bbox, diagramTitle);
+
+    const att = await this.uploadAttachmentBuffer(
+      pageId,
+      Buffer.from(svg, "utf-8"),
+      "diagram.drawio.svg",
+      "image/svg+xml",
+    );
+    const newSrc = `/api/files/${att.id}/${att.fileName}`;
+
+    const collabToken = await this.getCollabTokenWithReauth();
+    const pageUuid = await this.resolvePageId(pageId);
+
+    let repointed = 0;
+    const repoint = (n: any) => {
+      n.attrs = {
+        ...n.attrs,
+        src: newSrc,
+        attachmentId: att.id,
+        width: prepared.bbox.width,
+        height: prepared.bbox.height,
+      };
+      repointed++;
+    };
+
+    const mutation = await this.mutatePage(
+      pageUuid,
+      collabToken,
+      this.apiUrl,
+      (liveDoc) => {
+        repointed = 0;
+        const doc =
+          liveDoc && liveDoc.type === "doc"
+            ? liveDoc
+            : { type: "doc", content: [] };
+        if (!Array.isArray(doc.content)) doc.content = [];
+        // Repoint ONLY the resolved node — never every node that happens to
+        // share this attachmentId (a copied diagram is two nodes with one
+        // attachmentId; keying on it would clobber both). Re-resolve the same
+        // handle against the live doc and walk to its exact position.
+        const hit = getNodeByRef(doc, ref);
+        if (!hit || hit.type !== "drawio") return null; // vanished/changed -> skip
+        let target: any = doc;
+        for (const idx of hit.path) {
+          if (!target || !Array.isArray(target.content)) {
+            target = null;
+            break;
+          }
+          target = target.content[idx];
+        }
+        if (!target || target.type !== "drawio") return null;
+        repoint(target);
+        if (repointed === 0) return null; // node vanished concurrently -> skip
+        return doc;
+      },
+    );
+
+    if (repointed === 0) {
+      return {
+        success: true,
+        nodeId,
+        attachmentId: att.id,
+        warnings: [
+          ...prepared.warnings,
+          "target drawio node was removed concurrently; uploaded attachment is unreferenced",
+        ],
+        verify: mutation.verify,
+      };
+    }
+
+    return {
+      success: true,
+      nodeId,
+      attachmentId: att.id,
+      warnings: prepared.warnings,
+      verify: mutation.verify,
+    };
+  }
+
   // --- Page history / diff / transform ---
 
   /**

packages/mcp/src/index.ts

@@ -4,7 +4,7 @@ import { readFileSync } from "fs";
 import { fileURLToPath } from "url";
 import { dirname, join } from "path";
 import { DocmostClient, DocmostMcpConfig } from "./client.js";
-import { parseNodeArg } from "@docmost/prosemirror-markdown";
+import { parseNodeArg } from "./lib/parse-node-arg.js";
 import { SHARED_TOOL_SPECS, SharedToolSpec } from "./tool-specs.js";
 
 // Re-export the client and its config type so embedding hosts (e.g. the gitmost
@@ -47,7 +47,7 @@ const VERSION = packageJson.version;
 export const SERVER_INSTRUCTIONS =
   "Docmost editing guide — choose the tool by intent.\n" +
   "READ: find a page -> search (workspace-wide full-text); list -> list_pages / list_spaces. Locate blocks and their ids CHEAPLY -> get_outline (compact top-level map; start here, not get_page_json). One block's subtree -> get_node (by attrs.id, or \"#<index>\" for tables, which carry no id). Find every occurrence of a string/regex ON a page (and where each is) -> search_in_page, NOT block-by-block get_node — it returns each hit's node ref + block index + context for a targeted comment. Whole page -> get_page (Markdown, lossy; inline <span data-comment-id> tags are comment anchors — markup, not text) or get_page_json (lossless ProseMirror with block ids). Hand a huge page (with images) to an external consumer without pulling it through the model context -> stash_page (returns a short-lived anonymous URL).\n" +
-  "EDIT: fix wording/typos/numbers -> edit_page_text (find/replace inside blocks, no node id needed). Change ONE block (paragraph/heading/callout/etc.) structurally -> patch_node (by attrs.id from get_outline). Add a block -> insert_node (before/after a block by attrs.id or by anchor text, or append). Remove a block -> delete_node (by attrs.id). Tables -> table_get / table_update_cell / table_insert_row / table_delete_row (address by \"#<index>\" from get_outline; table nodes have no attrs.id). Images -> insert_image (add from a web URL) / replace_image (swap an existing image). Footnotes -> insert_footnote. Bulk/structural rewrite -> update_page_json (full ProseMirror replace; prefer the granular tools above to avoid resending the whole ~100KB+ document). Complex/scripted rewrite (multiple coordinated edits, renumbering) -> docmost_transform: write a JS `(doc, ctx) => doc` transform, preview the diff with dryRun (default), then apply with dryRun:false; ctx.helpers includes commentsToFootnotes for turning inline comments into numbered footnotes.\n" +
+  "EDIT: fix wording/typos/numbers -> edit_page_text (find/replace inside blocks, no node id needed). Change ONE block (paragraph/heading/callout/etc.) structurally -> patch_node (by attrs.id from get_outline). Add a block -> insert_node (before/after a block by attrs.id or by anchor text, or append). Remove a block -> delete_node (by attrs.id). Tables -> table_get / table_update_cell / table_insert_row / table_delete_row (address by \"#<index>\" from get_outline; table nodes have no attrs.id). Images -> insert_image (add from a web URL) / replace_image (swap an existing image). Draw.io diagrams -> drawio_create (create from mxGraph XML and insert), drawio_get (read a diagram as mxGraph XML + a hash), drawio_update (replace a diagram; pass the hash from drawio_get as baseHash for optimistic locking). Footnotes -> insert_footnote. Bulk/structural rewrite -> update_page_json (full ProseMirror replace; prefer the granular tools above to avoid resending the whole ~100KB+ document). Complex/scripted rewrite (multiple coordinated edits, renumbering) -> docmost_transform: write a JS `(doc, ctx) => doc` transform, preview the diff with dryRun (default), then apply with dryRun:false; ctx.helpers includes commentsToFootnotes for turning inline comments into numbered footnotes.\n" +
   "PAGES: new -> create_page (Markdown). Rename (title only) -> rename_page. Move -> move_page. Delete -> delete_page (SOFT delete — the page goes to trash and is restorable; nothing is permanent). Copy/replace a page's whole content from another page (server-side, no document through the model) -> copy_page_content. Sharing -> share_page / unshare_page / list_shares; share_page makes the page PUBLICLY accessible — do it only when explicitly asked.\n" +
   "COMMENTS: create_comment is always inline and requires an EXACT selection — contiguous text from a single block, <=250 chars (fails rather than leaving an unanchored comment); reply to a thread via parentCommentId. Propose a concrete text fix for one-click human approval -> create_comment with suggestedText (the exact plain-text replacement for the selection; the selection must then be UNIQUE in the page — extend it with context if needed); prefer this over editing directly when the change is subjective or needs the author's sign-off. Manage -> list_comments, update_comment, resolve_comment (resolve/reopen, reversible — prefer over delete to close), delete_comment, check_new_comments.\n" +
   "HISTORY: review what changed -> diff_page_versions (a historyId vs current, or two versions). List saved versions -> list_page_history. Undo a bad edit -> restore_page_version (writes a past version back as current; itself revertible). Lossless markdown round-trip (download, edit, re-upload, incl. comment anchors) -> export_page_markdown / import_page_markdown.";
@@ -460,6 +460,38 @@ registerShared(
   },
 );
 
+// Tool: drawio_get — read a draw.io diagram as mxGraph XML (or the raw SVG).
+registerShared(
+  SHARED_TOOL_SPECS.drawioGet,
+  async ({ pageId, node, format }) => {
+    const result = await docmostClient.drawioGet(pageId, node, format ?? "xml");
+    return jsonContent(result);
+  },
+);
+
+// Tool: drawio_create — lint mxGraph XML, build the .drawio.svg, insert a node.
+registerShared(
+  SHARED_TOOL_SPECS.drawioCreate,
+  async ({ pageId, xml, position, anchorNodeId, anchorText, title }) => {
+    const result = await docmostClient.drawioCreate(
+      pageId,
+      { position, anchorNodeId, anchorText },
+      xml,
+      title,
+    );
+    return jsonContent(result);
+  },
+);
+
+// Tool: drawio_update — optimistic-locked full replacement of a diagram.
+registerShared(
+  SHARED_TOOL_SPECS.drawioUpdate,
+  async ({ pageId, node, xml, baseHash }) => {
+    const result = await docmostClient.drawioUpdate(pageId, node, xml, baseHash);
+    return jsonContent(result);
+  },
+);
+
 // Tool: share_page
 // Schema + description now live in the shared registry (#294). The execute body
 // keeps this transport's own `searchIndexing ?? true` default.

packages/mcp/src/lib/collaboration.ts

@@ -14,7 +14,7 @@ import { JSDOM } from "jsdom";
 import { markdownToProseMirror } from "@docmost/prosemirror-markdown";
 import { docmostExtensions, docmostSchema } from "./docmost-schema.js";
 import { withPageLock } from "./page-lock.js";
-import { sanitizeForYjs, findUnstorableAttr } from "@docmost/prosemirror-markdown";
+import { sanitizeForYjs, findUnstorableAttr } from "./node-ops.js";
 import { canonicalizeFootnotes } from "./footnote-canonicalize.js";
 import { summarizeChange, VerifyReport } from "./diff.js";
 

packages/mcp/src/lib/drawio-preview.ts

@@ -0,0 +1,193 @@
+// Pure-TS schematic SVG preview for draw.io diagrams (issue #423, stage 1).
+//
+// HARD CONSTRAINT: no backend rendering. This is a dependency-free string
+// builder — given the parsed mxGraph cells it draws a rough schematic (rects,
+// ellipses, diamonds, edges + labels) that stands in as the diagram's visible
+// image UNTIL a human first opens it in the draw.io editor and saves, at which
+// point the client replaces this with the pixel-perfect export SVG. It is
+// deliberately approximate: it exists so a freshly-agent-created diagram is not
+// an empty box in the page.
+
+import type { DrawioCell, DrawioBBox } from "./drawio-xml.js";
+import { absolutePos } from "./drawio-xml.js";
+
+function esc(s: string): string {
+  return s
+    .replace(/&/g, "&")
+    .replace(/</g, "&lt;")
+    .replace(/>/g, "&gt;")
+    .replace(/"/g, "&quot;");
+}
+
+/**
+ * Strip HTML markup from a cell value (draw.io labels are HTML when html=1),
+ * decode the handful of entities we care about, and collapse whitespace so the
+ * label fits on the schematic. `<br>` becomes a space (this is a one-line
+ * preview label, not a faithful multi-line render).
+ */
+function labelText(value: string): string {
+  return value
+    .replace(/<br\s*\/?>/gi, " ")
+    .replace(/<[^>]+>/g, "")
+    .replace(/&lt;/g, "<")
+    .replace(/&gt;/g, ">")
+    .replace(/&quot;/g, '"')
+    .replace(/&#39;|&apos;/g, "'")
+    .replace(/&#xa;|&#10;/gi, " ")
+    .replace(/&amp;/g, "&")
+    .replace(/\s+/g, " ")
+    .trim();
+}
+
+function centeredLabel(cx: number, cy: number, value: string, color = "#000000"): string {
+  const text = labelText(value);
+  if (!text) return "";
+  return (
+    `<text x="${round(cx)}" y="${round(cy)}" ` +
+    `font-family="Helvetica, Arial, sans-serif" font-size="12" ` +
+    `text-anchor="middle" dominant-baseline="middle" fill="${esc(color)}">` +
+    `${esc(text)}</text>`
+  );
+}
+
+function round(n: number): number {
+  return Math.round(n * 100) / 100;
+}
+
+interface ShapeKind {
+  kind: "ellipse" | "rhombus" | "triangle" | "rect";
+}
+
+/**
+ * Decide which schematic primitive to draw for a vertex. A shape can be named
+ * either as the style's base token (e.g. "ellipse;…") or as a key (e.g.
+ * "shape=rhombus" / "ellipse=1"), so both the base style and the map are
+ * checked.
+ */
+function shapeKind(
+  styleMap: Record<string, string>,
+  baseStyle?: string,
+): ShapeKind {
+  const shape = styleMap.shape ?? baseStyle;
+  const has = (name: string) => shape === name || styleMap[name] != null;
+  if (has("ellipse")) return { kind: "ellipse" };
+  if (has("rhombus")) return { kind: "rhombus" };
+  if (has("triangle")) return { kind: "triangle" };
+  // Everything else — including unknown stencils (shape=mxgraph.*), swimlanes,
+  // and plain boxes — is drawn as a (rounded) rectangle.
+  return { kind: "rect" };
+}
+
+function fill(styleMap: Record<string, string>): string {
+  const c = styleMap.fillColor;
+  if (!c || c.toLowerCase() === "none") return "#ffffff";
+  return c;
+}
+
+function stroke(styleMap: Record<string, string>): string {
+  const c = styleMap.strokeColor;
+  if (!c || c.toLowerCase() === "none") return "#000000";
+  return c;
+}
+
+/**
+ * Render the schematic shapes as the INNER content of the `.drawio.svg` (the
+ * outer <svg> wrapper is added by drawio-xml.buildDrawioSvg). Coordinates are
+ * absolute (container children are resolved via the parent chain).
+ */
+export function renderDiagramShapes(cells: DrawioCell[], _bbox: DrawioBBox): string {
+  const byId = new Map(cells.map((c) => [c.id, c]));
+  const parts: string[] = [];
+
+  // Edges first so vertices sit on top of their connectors.
+  for (const c of cells) {
+    if (!c.edge) continue;
+    parts.push(renderEdge(c, byId));
+  }
+
+  for (const c of cells) {
+    if (!c.vertex || !c.geometry.hasGeometry) continue;
+    const g = c.geometry;
+    if (g.width == null || g.height == null) continue;
+    const { x, y } = absolutePos(c, byId);
+    parts.push(renderVertex(c, x, y, g.width, g.height));
+  }
+
+  return `<g>${parts.filter(Boolean).join("")}</g>`;
+}
+
+function renderVertex(
+  c: DrawioCell,
+  x: number,
+  y: number,
+  w: number,
+  h: number,
+): string {
+  const f = esc(fill(c.styleMap));
+  const s = esc(stroke(c.styleMap));
+  const { kind } = shapeKind(c.styleMap, c.baseStyle);
+  const cx = x + w / 2;
+  const cy = y + h / 2;
+  let shape = "";
+  switch (kind) {
+    case "ellipse":
+      shape =
+        `<ellipse cx="${round(cx)}" cy="${round(cy)}" rx="${round(w / 2)}" ` +
+        `ry="${round(h / 2)}" fill="${f}" stroke="${s}"/>`;
+      break;
+    case "rhombus": {
+      const pts = [
+        `${round(cx)},${round(y)}`,
+        `${round(x + w)},${round(cy)}`,
+        `${round(cx)},${round(y + h)}`,
+        `${round(x)},${round(cy)}`,
+      ].join(" ");
+      shape = `<polygon points="${pts}" fill="${f}" stroke="${s}"/>`;
+      break;
+    }
+    case "triangle": {
+      const pts = [
+        `${round(x)},${round(y)}`,
+        `${round(x + w)},${round(cy)}`,
+        `${round(x)},${round(y + h)}`,
+      ].join(" ");
+      shape = `<polygon points="${pts}" fill="${f}" stroke="${s}"/>`;
+      break;
+    }
+    default: {
+      const rounded = c.styleMap.rounded === "1";
+      const rx = rounded ? Math.min(12, w / 2, h / 2) : 0;
+      shape =
+        `<rect x="${round(x)}" y="${round(y)}" width="${round(w)}" ` +
+        `height="${round(h)}" rx="${round(rx)}" ry="${round(rx)}" ` +
+        `fill="${f}" stroke="${s}"/>`;
+    }
+  }
+  return shape + centeredLabel(cx, cy, c.value, c.styleMap.fontColor || "#000000");
+}
+
+function renderEdge(c: DrawioCell, byId: Map<string, DrawioCell>): string {
+  const src = c.source != null ? byId.get(c.source) : undefined;
+  const tgt = c.target != null ? byId.get(c.target) : undefined;
+  const p1 = anchorPoint(src, byId);
+  const p2 = anchorPoint(tgt, byId);
+  if (!p1 || !p2) return ""; // a floating endpoint with no fixed point: skip
+  const line =
+    `<line x1="${round(p1.x)}" y1="${round(p1.y)}" ` +
+    `x2="${round(p2.x)}" y2="${round(p2.y)}" ` +
+    `stroke="#000000" stroke-width="1"/>`;
+  const mid = { x: (p1.x + p2.x) / 2, y: (p1.y + p2.y) / 2 };
+  return line + centeredLabel(mid.x, mid.y, c.value);
+}
+
+/** Center point of a vertex used as an edge anchor (approximate). */
+function anchorPoint(
+  cell: DrawioCell | undefined,
+  byId: Map<string, DrawioCell>,
+): { x: number; y: number } | null {
+  if (!cell || !cell.geometry.hasGeometry) return null;
+  const g = cell.geometry;
+  if (g.width == null || g.height == null) return null;
+  const { x, y } = absolutePos(cell, byId);
+  return { x: x + g.width / 2, y: y + g.height / 2 };
+}

packages/mcp/src/lib/drawio-xml.ts

@@ -0,0 +1,771 @@
+// draw.io (mxGraph) XML support for the MCP drawio tools (issue #423, stage 1).
+//
+// This module owns everything that is pure data-plumbing for draw.io diagrams:
+//   - the DECODE CHAIN that turns a stored `diagram.drawio.svg` attachment back
+//     into mxGraph XML (handles both the plain nested-XML form Docmost writes
+//     and draw.io's own COMPRESSED `<diagram>` payload — base64 + raw-deflate);
+//   - the ENCODE side that wraps mxGraph XML into the `.drawio.svg` attachment
+//     using the exact same contract as the import service's createDrawioSvg;
+//   - a deterministic LINTER that rejects the structural mistakes generators
+//     make before anything is written (each violation carries the offending
+//     cellId + position so the model can auto-retry);
+//   - a stable HASH over the normalized XML, used as the optimistic-lock key.
+//
+// HARD CONSTRAINT: no backend rendering. Nothing here shells out or renders a
+// bitmap; the only runtime dependencies are jsdom (already used across this
+// package for XML parsing) and pako (raw-inflate for the compressed format).
+
+import { createHash } from "node:crypto";
+import { JSDOM } from "jsdom";
+import pako from "pako";
+
+// --- shared XML parser -----------------------------------------------------
+
+// A single reusable JSDOM window; constructing one per parse is wasteful and
+// these tools are low-frequency. Only the DOMParser is used.
+let _window: any = null;
+function xmlWindow(): any {
+  if (!_window) _window = new JSDOM("").window;
+  return _window;
+}
+
+/** Default mxGraphModel attributes used when the server wraps a cell list. */
+const DEFAULT_MODEL_ATTRS =
+  'dx="0" dy="0" grid="1" gridSize="10" page="1" pageWidth="850" pageHeight="1100"';
+
+// --- structured lint errors ------------------------------------------------
+
+export interface DrawioLintIssue {
+  /** Machine-readable rule id, e.g. "edge-geometry". */
+  rule: string;
+  /** Human-readable explanation the model can act on. */
+  message: string;
+  /** The offending cell's id, when the rule is cell-scoped. */
+  cellId?: string;
+  /** Extra location info: cell index in <root>, or a parser line:col. */
+  position?: string;
+}
+
+/**
+ * Thrown by the linter and by decode/prepare when the input is unusable. Carries
+ * the full list of issues so the caller can surface a structured tool-error the
+ * model auto-retries against.
+ */
+export class DrawioLintError extends Error {
+  issues: DrawioLintIssue[];
+  constructor(issues: DrawioLintIssue[]) {
+    const summary = issues
+      .map((i) => {
+        const where = [
+          i.cellId != null ? `cellId=${i.cellId}` : null,
+          i.position != null ? `at ${i.position}` : null,
+        ]
+          .filter(Boolean)
+          .join(", ");
+        return `[${i.rule}] ${i.message}${where ? ` (${where})` : ""}`;
+      })
+      .join("; ");
+    super(`drawio lint failed: ${summary}`);
+    this.name = "DrawioLintError";
+    this.issues = issues;
+  }
+}
+
+// --- parsed-cell model -----------------------------------------------------
+
+export interface DrawioGeometry {
+  x?: number;
+  y?: number;
+  width?: number;
+  height?: number;
+  relative: boolean;
+  hasGeometry: boolean;
+}
+
+export interface DrawioCell {
+  id: string;
+  parent?: string;
+  source?: string;
+  target?: string;
+  vertex: boolean;
+  edge: boolean;
+  value: string;
+  style: string;
+  styleMap: Record<string, string>;
+  /** Non-key/value leading token of the style (a base stylename), if any. */
+  baseStyle?: string;
+  geometry: DrawioGeometry;
+}
+
+export interface DrawioBBox {
+  width: number;
+  height: number;
+}
+
+// --- style parsing ---------------------------------------------------------
+
+/**
+ * Parse a draw.io style string into { baseStyle, map }. Grammar:
+ *   [stylename;]key=value;key=value;...
+ * A single leading token without '=' is the base stylename (e.g. "text" or
+ * "ellipse"). Every other non-empty segment must be exactly one key=value pair.
+ * Returns `null` (the segment index) on the first malformed segment so the
+ * linter can report a precise error.
+ */
+export function parseStyle(
+  style: string,
+): { baseStyle?: string; map: Record<string, string>; badSegment?: string } {
+  const map: Record<string, string> = {};
+  let baseStyle: string | undefined;
+  const segments = style.split(";");
+  for (let i = 0; i < segments.length; i++) {
+    const seg = segments[i].trim();
+    if (seg === "") continue; // trailing/empty segments are fine
+    const eq = seg.indexOf("=");
+    if (eq === -1) {
+      // A bare token is only valid as the FIRST meaningful segment (base style).
+      if (baseStyle === undefined && Object.keys(map).length === 0) {
+        baseStyle = seg;
+        continue;
+      }
+      return { baseStyle, map, badSegment: seg };
+    }
+    // A second '=' inside the same segment is malformed.
+    if (seg.indexOf("=", eq + 1) !== -1) {
+      return { baseStyle, map, badSegment: seg };
+    }
+    const key = seg.slice(0, eq).trim();
+    const val = seg.slice(eq + 1).trim();
+    if (key === "") return { baseStyle, map, badSegment: seg };
+    map[key] = val;
+  }
+  return { baseStyle, map };
+}
+
+// --- low-level XML helpers -------------------------------------------------
+
+function parseXml(xml: string): { doc: any; error: string | null } {
+  const parser = new (xmlWindow().DOMParser)();
+  const doc = parser.parseFromString(xml, "application/xml");
+  const err = doc.getElementsByTagName("parsererror");
+  if (err.length > 0) {
+    // jsdom prefixes the message with "line:col:" — keep it as the position.
+    return { doc, error: (err[0].textContent || "malformed XML").trim() };
+  }
+  return { doc, error: null };
+}
+
+function num(v: string | null): number | undefined {
+  if (v == null || v === "") return undefined;
+  const n = Number(v);
+  return Number.isFinite(n) ? n : undefined;
+}
+
+/** Extract the raw `<mxGraphModel …>…</mxGraphModel>` substring, or null. */
+function sliceModel(xml: string): string | null {
+  const open = xml.indexOf("<mxGraphModel");
+  if (open === -1) return null;
+  const close = xml.indexOf("</mxGraphModel>", open);
+  if (close === -1) {
+    // Self-closed empty model, e.g. `<mxGraphModel .../>`.
+    const selfClose = xml.indexOf("/>", open);
+    if (selfClose !== -1) return xml.slice(open, selfClose + 2);
+    return null;
+  }
+  return xml.slice(open, close + "</mxGraphModel>".length);
+}
+
+// --- decode chain ----------------------------------------------------------
+
+/**
+ * Read the `content=` attribute out of a `.drawio.svg` string. Docmost stores a
+ * base64 payload there (createDrawioSvg); draw.io's own SVG export may store the
+ * XML entity-encoded instead. The DOM decodes entities for us, so the caller
+ * only has to distinguish "starts with '<'" (raw XML) from base64.
+ */
+export function extractContentAttr(svg: string): string {
+  const { doc, error } = parseXml(svg);
+  if (!error) {
+    const root = doc.documentElement;
+    if (root && root.hasAttribute && root.hasAttribute("content")) {
+      return root.getAttribute("content") || "";
+    }
+  }
+  // Fallback for a malformed wrapper: pull the attribute directly. The content
+  // value itself never contains a double-quote (base64 / entity-encoded XML).
+  const m = /content="([^"]*)"/.exec(svg);
+  if (m) {
+    // Decode the handful of XML entities a raw regex would leave encoded.
+    return m[1]
+      .replace(/&lt;/g, "<")
+      .replace(/&gt;/g, ">")
+      .replace(/&quot;/g, '"')
+      .replace(/&#39;/g, "'")
+      .replace(/&amp;/g, "&");
+  }
+  throw new Error("drawio: SVG has no content= attribute to decode");
+}
+
+/**
+ * Turn a decoded draw.io file (`<mxfile>` or a bare `<mxGraphModel>`, possibly
+ * with a COMPRESSED `<diagram>` payload) into the mxGraphModel XML. For the
+ * plain form the raw substring is returned verbatim so a round-trip stays
+ * byte-stable; the compressed form is inflated (base64 → raw-deflate →
+ * decodeURIComponent), which is how draw.io stores diagrams by default.
+ */
+export function decodeDrawioFileToModel(fileXml: string): string {
+  // Plain, nested XML: return the model substring untouched (byte-stable).
+  const sliced = sliceModel(fileXml);
+  if (sliced) return sliced;
+
+  // Otherwise it must be the compressed `<diagram>…</diagram>` text payload.
+  const open = fileXml.indexOf("<diagram");
+  if (open !== -1) {
+    const gt = fileXml.indexOf(">", open);
+    const close = fileXml.indexOf("</diagram>", gt);
+    if (gt !== -1 && close !== -1) {
+      const payload = fileXml.slice(gt + 1, close).trim();
+      if (payload) {
+        const inflated = inflateDiagramPayload(payload);
+        const model = sliceModel(inflated);
+        if (model) return model;
+        return inflated;
+      }
+    }
+  }
+  throw new Error(
+    "drawio: could not decode file — no <mxGraphModel> and no compressed <diagram> payload",
+  );
+}
+
+/**
+ * Upper bound on the inflated size of a compressed `<diagram>` payload
+ * (decompression-bomb guard). `fetchInternalFile` caps the DOWNLOAD at 64 MiB,
+ * but a tiny crafted compressed payload can inflate to gigabytes and OOM the
+ * process. A real diagram's mxGraphModel XML is small (KBs to low MBs even for
+ * large diagrams), so 16 MiB is far above any legitimate payload while keeping
+ * memory bounded. Chars ~= bytes for the (mostly ASCII) URI-encoded XML.
+ */
+export const MAX_INFLATED_DIAGRAM_BYTES = 16 * 1024 * 1024;
+
+/**
+ * Inflate draw.io's compressed diagram payload:
+ *   base64-decode → raw-inflate (raw deflate, windowBits -15) →
+ *   decodeURIComponent.
+ *
+ * Uses pako's streaming Inflate so we can abort as soon as the decompressed
+ * output exceeds MAX_INFLATED_DIAGRAM_BYTES — the full bomb is never
+ * materialised in memory.
+ */
+export function inflateDiagramPayload(base64: string): string {
+  const bytes = Buffer.from(base64, "base64");
+  const inflator = new pako.Inflate({ raw: true, to: "string" });
+  let total = 0;
+  const passthrough = inflator.onData.bind(inflator);
+  inflator.onData = (chunk: string | Uint8Array) => {
+    total += chunk.length;
+    if (total > MAX_INFLATED_DIAGRAM_BYTES) {
+      // Throwing here propagates out of push(), aborting inflation immediately.
+      throw new Error(
+        `drawio: refusing to decode diagram — decompressed size exceeds ` +
+          `${MAX_INFLATED_DIAGRAM_BYTES} bytes (possible decompression bomb)`,
+      );
+    }
+    passthrough(chunk);
+  };
+  inflator.push(bytes, true);
+  if (inflator.err) {
+    throw new Error(
+      `drawio: failed to inflate compressed <diagram> payload (${inflator.msg || inflator.err})`,
+    );
+  }
+  const uriEncoded = inflator.result as string;
+  return decodeURIComponent(uriEncoded);
+}
+
+/** Full decode chain: `.drawio.svg` string → mxGraphModel XML. */
+export function decodeDrawioSvg(svg: string): string {
+  const content = extractContentAttr(svg).trim();
+  const fileXml = content.startsWith("<")
+    ? content
+    : Buffer.from(content, "base64").toString("utf-8");
+  return decodeDrawioFileToModel(fileXml);
+}
+
+// --- encode side -----------------------------------------------------------
+
+/**
+ * Wrap an mxGraphModel in the plain (uncompressed) `<mxfile><diagram>` envelope.
+ * draw.io opens uncompressed XML fine, and staying uncompressed keeps the
+ * write path deterministic and the round-trip byte-stable.
+ */
+export function encodeDrawioFile(modelXml: string, title = "Page-1"): string {
+  const safeTitle = xmlEscape(title);
+  return `<mxfile host="drawio"><diagram id="page-1" name="${safeTitle}">${modelXml}</diagram></mxfile>`;
+}
+
+/**
+ * Build the `diagram.drawio.svg` attachment. Mirrors the import service's
+ * createDrawioSvg contract exactly:
+ *   <svg xmlns=… xmlns:xlink=… content="${base64(drawioFile)}">${inner}</svg>
+ * plus width/height/viewBox from the diagram bounding box and the schematic
+ * preview as the visible children (`inner`).
+ */
+export function buildDrawioSvg(
+  modelXml: string,
+  inner: string,
+  bbox: DrawioBBox,
+  title = "Page-1",
+): string {
+  const file = encodeDrawioFile(modelXml, title);
+  const base64 = Buffer.from(file, "utf-8").toString("base64");
+  const w = Math.max(1, Math.round(bbox.width));
+  const h = Math.max(1, Math.round(bbox.height));
+  return (
+    `<svg xmlns="http://www.w3.org/2000/svg" ` +
+    `xmlns:xlink="http://www.w3.org/1999/xlink" ` +
+    `width="${w}" height="${h}" viewBox="0 0 ${w} ${h}" ` +
+    `content="${base64}">${inner}</svg>`
+  );
+}
+
+function xmlEscape(s: string): string {
+  return s
+    .replace(/&/g, "&amp;")
+    .replace(/</g, "&lt;")
+    .replace(/>/g, "&gt;")
+    .replace(/"/g, "&quot;");
+}
+
+// --- normalization + hash --------------------------------------------------
+
+/**
+ * Normalize mxGraph XML for hashing / stable comparison: drop the whitespace
+ * between tags and trim. This is intentionally conservative — it never reorders
+ * attributes or cells (that would be lossy) — so two documents hash equal iff
+ * they differ only in inter-tag formatting.
+ */
+export function normalizeXml(xml: string): string {
+  return xml.replace(/>\s+</g, "><").trim();
+}
+
+/** Stable optimistic-lock hash over the normalized model XML (sha256, hex). */
+export function mxHash(modelXml: string): string {
+  return createHash("sha256").update(normalizeXml(modelXml), "utf-8").digest("hex");
+}
+
+// --- cell parsing ----------------------------------------------------------
+
+/** Parse every `<mxCell>` in a model into a structured DrawioCell list. */
+export function parseCells(modelXml: string): DrawioCell[] {
+  const { doc, error } = parseXml(modelXml);
+  if (error) {
+    throw new DrawioLintError([
+      { rule: "well-formed-xml", message: error, position: firstLineCol(error) },
+    ]);
+  }
+  const cells: DrawioCell[] = [];
+  const els = doc.getElementsByTagName("mxCell");
+  for (let i = 0; i < els.length; i++) {
+    cells.push(readCell(els[i]));
+  }
+  return cells;
+}
+
+function readCell(el: any): DrawioCell {
+  const style = el.getAttribute("style") || "";
+  const parsed = parseStyle(style);
+  const geoEl = firstChildByTag(el, "mxGeometry");
+  const geometry: DrawioGeometry = geoEl
+    ? {
+        x: num(geoEl.getAttribute("x")),
+        y: num(geoEl.getAttribute("y")),
+        width: num(geoEl.getAttribute("width")),
+        height: num(geoEl.getAttribute("height")),
+        relative: geoEl.getAttribute("relative") === "1",
+        hasGeometry: true,
+      }
+    : { relative: false, hasGeometry: false };
+  return {
+    id: el.getAttribute("id") ?? "",
+    parent: el.getAttribute("parent") ?? undefined,
+    source: el.getAttribute("source") ?? undefined,
+    target: el.getAttribute("target") ?? undefined,
+    vertex: el.getAttribute("vertex") === "1",
+    edge: el.getAttribute("edge") === "1",
+    value: el.getAttribute("value") ?? "",
+    style,
+    styleMap: parsed.map,
+    baseStyle: parsed.baseStyle,
+    geometry,
+  };
+}
+
+function firstChildByTag(el: any, tag: string): any {
+  for (let i = 0; i < el.childNodes.length; i++) {
+    const c = el.childNodes[i];
+    if (c.nodeType === 1 && c.tagName === tag) return c;
+  }
+  return null;
+}
+
+function firstLineCol(msg: string): string | undefined {
+  const m = /^(\d+:\d+)/.exec(msg);
+  return m ? m[1] : undefined;
+}
+
+// --- bounding box ----------------------------------------------------------
+
+/**
+ * Absolute bounding box of the diagram from its vertex geometries. Container
+ * children are relative, so absolute positions are resolved along the parent
+ * chain before taking the extent. Falls back to a default canvas when empty.
+ */
+export function computeBBox(cells: DrawioCell[]): DrawioBBox {
+  const byId = new Map(cells.map((c) => [c.id, c]));
+  let maxX = 0;
+  let maxY = 0;
+  let any = false;
+  for (const c of cells) {
+    if (!c.vertex || !c.geometry.hasGeometry) continue;
+    const g = c.geometry;
+    if (g.width == null || g.height == null) continue;
+    const { x, y } = absolutePos(c, byId);
+    maxX = Math.max(maxX, x + g.width);
+    maxY = Math.max(maxY, y + g.height);
+    any = true;
+  }
+  if (!any) return { width: 300, height: 200 };
+  // A small margin so borders/labels are not clipped at the edge.
+  return { width: Math.ceil(maxX) + 20, height: Math.ceil(maxY) + 20 };
+}
+
+/** Absolute (x,y) of a vertex, following its parent chain (containers). */
+export function absolutePos(
+  cell: DrawioCell,
+  byId: Map<string, DrawioCell>,
+): { x: number; y: number } {
+  let x = cell.geometry.x ?? 0;
+  let y = cell.geometry.y ?? 0;
+  const seen = new Set<string>([cell.id]);
+  let parentId = cell.parent;
+  while (parentId && !seen.has(parentId)) {
+    seen.add(parentId);
+    const p = byId.get(parentId);
+    // Sentinels (0/1) carry no geometry; stop there.
+    if (!p || !p.vertex || !p.geometry.hasGeometry) break;
+    x += p.geometry.x ?? 0;
+    y += p.geometry.y ?? 0;
+    parentId = p.parent;
+  }
+  return { x, y };
+}
+
+// --- linter ----------------------------------------------------------------
+
+/**
+ * Run every deterministic pre-write rule over a full mxGraphModel string. On any
+ * violation it throws a DrawioLintError carrying one issue per violation, each
+ * with the offending cellId + position. Returns the parsed cells on success.
+ */
+export function lintModel(modelXml: string): {
+  cells: DrawioCell[];
+  warnings: string[];
+} {
+  const issues: DrawioLintIssue[] = [];
+  const warnings: string[] = [];
+
+  // Rule: no XML comments. Checked on the raw string (a comment survives DOM
+  // parsing as a comment node, but the intent is to reject them outright — they
+  // routinely wrap "TODO" cruft that breaks downstream tooling).
+  if (modelXml.includes("<!--")) {
+    issues.push({
+      rule: "no-comments",
+      message: "XML comments (<!-- -->) are not allowed in diagram XML",
+    });
+  }
+
+  // Rule: value escaping + literal newline. Scan raw <mxCell> tags so the error
+  // can name the cell id even when the whole document is otherwise malformed.
+  scanRawValues(modelXml, issues);
+
+  // Well-formedness — everything below needs a parsed DOM.
+  const { doc, error } = parseXml(modelXml);
+  if (error) {
+    issues.push({
+      rule: "well-formed-xml",
+      message: error,
+      position: firstLineCol(error),
+    });
+    throw new DrawioLintError(issues);
+  }
+
+  const root = doc.documentElement;
+  if (!root || root.tagName !== "mxGraphModel") {
+    issues.push({
+      rule: "structure",
+      message: `root element must be <mxGraphModel>, got <${root ? root.tagName : "?"}>`,
+    });
+    throw new DrawioLintError(issues);
+  }
+  if (!firstChildByTag(root, "root")) {
+    issues.push({
+      rule: "structure",
+      message: "<mxGraphModel> must contain a <root> element",
+    });
+    throw new DrawioLintError(issues);
+  }
+
+  const cells = parseCells(modelXml);
+  const ids = new Set<string>();
+
+  // Rule: sentinel cells id="0" and id="1"(parent="0").
+  const cell0 = cells.find((c) => c.id === "0");
+  const cell1 = cells.find((c) => c.id === "1");
+  if (!cell0) {
+    issues.push({
+      rule: "sentinel-cells",
+      message: 'missing the root sentinel cell <mxCell id="0"/>',
+      cellId: "0",
+    });
+  }
+  if (!cell1) {
+    issues.push({
+      rule: "sentinel-cells",
+      message: 'missing the layer sentinel cell <mxCell id="1" parent="0"/>',
+      cellId: "1",
+    });
+  } else if (cell1.parent !== "0") {
+    issues.push({
+      rule: "sentinel-cells",
+      message: 'the layer sentinel <mxCell id="1"> must have parent="0"',
+      cellId: "1",
+    });
+  }
+
+  cells.forEach((c, index) => {
+    const pos = `cell #${index}`;
+    const isSentinel = c.id === "0" || c.id === "1";
+
+    // Rule: unique, non-empty ids; user cells must not reuse 0/1.
+    if (c.id === "") {
+      issues.push({ rule: "cell-id", message: "cell has an empty id", position: pos });
+    } else if (ids.has(c.id)) {
+      issues.push({
+        rule: "duplicate-id",
+        message: `duplicate cell id "${c.id}"`,
+        cellId: c.id,
+        position: pos,
+      });
+    }
+    ids.add(c.id);
+
+    if (isSentinel) return; // sentinels are exempt from the shape rules below
+
+    // Rule: vertex XOR edge (a cell may be neither: groups/containers).
+    if (c.vertex && c.edge) {
+      issues.push({
+        rule: "vertex-edge-exclusive",
+        message: 'a cell cannot be both vertex="1" and edge="1"',
+        cellId: c.id,
+        position: pos,
+      });
+    }
+
+    // Rule: every edge has a child <mxGeometry as="geometry"/>.
+    if (c.edge && !c.geometry.hasGeometry) {
+      issues.push({
+        rule: "edge-geometry",
+        message:
+          'edge is missing its child <mxGeometry relative="1" as="geometry"/> — it will not render',
+        cellId: c.id,
+        position: pos,
+      });
+    }
+
+    // Rule: edge endpoints resolve to existing ids.
+    if (c.edge) {
+      for (const end of ["source", "target"] as const) {
+        const ref = c[end];
+        if (ref != null && ref !== "" && !cellExists(cells, ref)) {
+          issues.push({
+            rule: "edge-endpoint",
+            message: `edge ${end} "${ref}" does not resolve to any cell`,
+            cellId: c.id,
+            position: pos,
+          });
+        }
+      }
+    }
+
+    // Rule: parent must exist.
+    if (c.parent != null && c.parent !== "" && !cellExists(cells, c.parent)) {
+      issues.push({
+        rule: "parent-exists",
+        message: `parent "${c.parent}" does not resolve to any cell`,
+        cellId: c.id,
+        position: pos,
+      });
+    }
+
+    // Rule: style parses as key=value; pairs.
+    if (c.style !== "") {
+      const parsed = parseStyle(c.style);
+      if (parsed.badSegment !== undefined) {
+        issues.push({
+          rule: "style-format",
+          message: `malformed style segment "${parsed.badSegment}" (expected key=value)`,
+          cellId: c.id,
+          position: pos,
+        });
+      }
+    }
+  });
+
+  if (issues.length > 0) throw new DrawioLintError(issues);
+  return { cells, warnings };
+}
+
+function cellExists(cells: DrawioCell[], id: string): boolean {
+  return cells.some((c) => c.id === id);
+}
+
+/**
+ * Raw-string scan of every `value="…"`/`value='…'` on an mxCell tag. Catches an
+ * unescaped `&`/`<`/`>` and a literal newline character inside a value, keyed to
+ * the cell's id. Runs before DOM parsing so a value bug is reported with its
+ * cellId even when the document is otherwise malformed.
+ */
+function scanRawValues(xml: string, issues: DrawioLintIssue[]): void {
+  const tagRe = /<mxCell\b([^>]*?)\/?>/g;
+  let m: RegExpExecArray | null;
+  while ((m = tagRe.exec(xml)) !== null) {
+    const attrs = m[1];
+    const idM = /\bid\s*=\s*"([^"]*)"/.exec(attrs);
+    const cellId = idM ? idM[1] : undefined;
+    const valM = /\bvalue\s*=\s*"([^"]*)"/.exec(attrs) || /\bvalue\s*=\s*'([^']*)'/.exec(attrs);
+    if (!valM) continue;
+    const raw = valM[1];
+    // Literal newline (0x0A / 0x0D) inside the attribute value.
+    if (/[\n\r]/.test(raw)) {
+      issues.push({
+        rule: "value-newline",
+        message:
+          "value contains a literal newline; use &#xa; (or <br> with html=1) instead",
+        cellId,
+      });
+    }
+    // Unescaped '<' or '>' inside a value.
+    if (raw.includes("<") || raw.includes(">")) {
+      issues.push({
+        rule: "value-escaping",
+        message: "value contains an unescaped '<' or '>'; use &lt; / &gt;",
+        cellId,
+      });
+    }
+    // '&' that does not begin a valid entity.
+    const badAmp = /&(?!(amp|lt|gt|quot|apos|#[0-9]+|#x[0-9a-fA-F]+);)/.test(raw);
+    if (badAmp) {
+      issues.push({
+        rule: "value-escaping",
+        message: "value contains an unescaped '&'; use &amp;",
+        cellId,
+      });
+    }
+  }
+}
+
+// --- input normalization + prepare -----------------------------------------
+
+/**
+ * Normalize an accepted tool input into a full mxGraphModel string:
+ *   - a bare `<mxGraphModel>` is used as-is;
+ *   - an `<mxfile>` is decoded to its first page's model;
+ *   - a list of `<mxCell>` is wrapped with the mxGraphModel/root envelope and
+ *     the sentinel cells (id=0, id=1 parent=0) are added when absent.
+ */
+export function normalizeInput(inputXml: string): string {
+  let xml = inputXml.trim();
+  // Strip an optional XML prolog.
+  if (xml.startsWith("<?xml")) {
+    const end = xml.indexOf("?>");
+    if (end !== -1) xml = xml.slice(end + 2).trim();
+  }
+  if (xml.startsWith("<mxfile")) {
+    return decodeDrawioFileToModel(xml);
+  }
+  if (xml.startsWith("<mxGraphModel")) {
+    return xml;
+  }
+  if (xml.includes("<mxCell")) {
+    return wrapCellFragment(xml);
+  }
+  throw new DrawioLintError([
+    {
+      rule: "unrecognized-input",
+      message:
+        "input must be a <mxGraphModel>, an <mxfile>, or a list of <mxCell> elements",
+    },
+  ]);
+}
+
+function wrapCellFragment(fragment: string): string {
+  // Validate the fragment is well-formed (wrapped so a bare list parses) and
+  // discover which sentinels are already present.
+  const { doc, error } = parseXml(`<root>${fragment}</root>`);
+  if (error) {
+    throw new DrawioLintError([
+      {
+        rule: "well-formed-xml",
+        message: error,
+        position: firstLineCol(error),
+      },
+    ]);
+  }
+  const existing = new Set<string>();
+  const els = doc.getElementsByTagName("mxCell");
+  for (let i = 0; i < els.length; i++) {
+    existing.add(els[i].getAttribute("id") ?? "");
+  }
+  let prefix = "";
+  if (!existing.has("0")) prefix += '<mxCell id="0"/>';
+  if (!existing.has("1")) prefix += '<mxCell id="1" parent="0"/>';
+  return `<mxGraphModel ${DEFAULT_MODEL_ATTRS}><root>${prefix}${fragment}</root></mxGraphModel>`;
+}
+
+export interface PreparedModel {
+  /** Canonical (normalized) mxGraphModel XML that gets written. */
+  modelXml: string;
+  cells: DrawioCell[];
+  bbox: DrawioBBox;
+  /** Number of user cells (excludes the id=0/id=1 sentinels). */
+  cellCount: number;
+  warnings: string[];
+  hash: string;
+}
+
+/**
+ * Full pre-write pipeline for create/update: normalize the input into a model,
+ * lint it (throws DrawioLintError on any violation), then compute the canonical
+ * form, bounding box, cell count and hash. Never touches the network.
+ */
+export function prepareModel(inputXml: string): PreparedModel {
+  const rawModel = normalizeInput(inputXml);
+  const { cells, warnings } = lintModel(rawModel);
+  const modelXml = normalizeXml(rawModel);
+  const bbox = computeBBox(cells);
+  const cellCount = cells.filter((c) => c.id !== "0" && c.id !== "1").length;
+  return {
+    modelXml,
+    cells,
+    bbox,
+    cellCount,
+    warnings,
+    hash: mxHash(modelXml),
+  };
+}
+
+/** Cell count of a decoded model (user cells only) — used by drawio_get meta. */
+export function countUserCells(modelXml: string): number {
+  return parseCells(modelXml).filter((c) => c.id !== "0" && c.id !== "1").length;
+}

packages/mcp/src/lib/footnote-analyze.ts

@@ -1,62 +1,136 @@
 /**
- * Legacy footnote advisory for imported Markdown (issue #166, reduced in #414).
+ * Legacy footnote diagnostics for imported Markdown (issue #166).
  *
- * Since #293 STEP 5 the canonical import form is inline `^[body]` footnotes
- * (handled by `@docmost/prosemirror-markdown`). LEGACY reference-style
- * `[^id]: …` definition markup is now INERT on import — the importer leaves it as
- * literal text — so authoring it silently produces broken footnotes (the #410
- * incident class). Rather than the old, elaborate diagnostics of every problem
- * SHAPE (dangling/duplicate/empty/in-table) that no longer describe what the
- * importer builds, this module surfaces ONE advisory warning whenever legacy
- * reference-style definition syntax is present, nudging the author to the inline
- * form. It never changes the document — the importer still creates the page.
+ * A PURE, fence-aware text scan (independent of the Markdown->ProseMirror
+ * conversion path, so it reports the same problems for `create_page`,
+ * `update_page` and `import_page_markdown`). It never changes the document — the
+ * importer still creates the page; this only surfaces footnote problems to the
+ * caller so an agent can fix its own markup instead of shipping broken footnotes.
  *
- * The scan is fence-aware: a `[^id]:` line inside a ``` / ~~~ code block is
- * example text, not markup, so it never triggers the warning.
+ * SCOPE after #293 STEP 5: the canonical import form is now inline `^[body]`
+ * footnotes (handled by `@docmost/prosemirror-markdown`), where these problems
+ * cannot arise. This scan therefore targets the LEGACY reference-style
+ * (`[^id]` / `[^id]:`) markup, which is now inert on import (left as literal
+ * text). The warnings remain useful as an advisory nudge when an agent still
+ * authors the old syntax, but they no longer describe what the importer builds.
+ *
+ * Detected problems:
+ *  - danglingReferences: a `[^id]` reference with no `[^id]:` definition.
+ *  - emptyDefinitions:   a `[^id]:` whose (kept) text is empty/whitespace.
+ *  - duplicateDefinitions: an id defined by two or more `[^id]:` lines (only the
+ *    first would have been kept under the old first-wins import).
+ *  - referencesInTables: a `[^id]` marker found in a GFM table row (heuristic:
+ *    the line, trimmed, starts with `|`) — footnotes in table cells often do not
+ *    render as expected.
  */
 
-/** A legacy footnote DEFINITION line: `[^id]:` at the start of a (non-fenced) line. */
-const FOOTNOTE_DEF_RE = /^\[\^[^\]\s]+\]:/;
-/** Opening/closing code fence marker (``` or ~~~). */
-const FENCE_RE = /^\s*(`{3,}|~{3,})/;
+import {
+  lexFootnoteLines,
+  forEachFootnoteReference,
+} from "./footnote-lex.js";
 
-/** The single advisory shown when legacy reference-style footnotes are present. */
-export const LEGACY_FOOTNOTE_WARNING =
-  "Reference-style footnotes (`[^id]: …`) are not parsed on import and will " +
-  "appear as literal text. Use inline footnotes instead: `^[footnote text]`.";
+export interface FootnoteDiagnostics {
+  /** Reference ids (distinct, document order) with no matching definition. */
+  danglingReferences: string[];
+  /** Definition ids whose first (kept) text is empty/whitespace. */
+  emptyDefinitions: string[];
+  /** Ids defined by two or more `[^id]:` lines (only the first is kept). */
+  duplicateDefinitions: string[];
+  /** Reference ids found inside a GFM table row (heuristic). */
+  referencesInTables: string[];
+  /** Human-readable warning lines for the tool result (one per problem class). */
+  warnings: string[];
+}
 
 /**
- * True when `markdown` contains a legacy `[^id]:` definition line OUTSIDE any
- * code fence. Pure; safe to call on any body.
+ * Analyze the footnotes in a Markdown string. Pure; safe to call on any body.
  */
-export function hasLegacyFootnoteDefinition(markdown: string): boolean {
-  if (typeof markdown !== "string" || !markdown.includes("[^")) return false;
-  let fence: string | null = null;
-  for (const line of markdown.split("\n")) {
-    const fenceMatch = FENCE_RE.exec(line);
-    if (fenceMatch) {
-      const marker = fenceMatch[1][0];
-      if (fence === null) fence = marker; // opening fence
-      else if (marker === fence) fence = null; // matching closing fence
+export function analyzeFootnotes(markdown: string): FootnoteDiagnostics {
+  // Distinct reference ids in first-appearance order, plus the set of ids seen
+  // inside a table row.
+  const refIds: string[] = [];
+  const refIdSet = new Set<string>();
+  const referencesInTables = new Set<string>();
+  const addRef = (id: string, inTable: boolean) => {
+    if (!refIdSet.has(id)) {
+      refIdSet.add(id);
+      refIds.push(id);
+    }
+    if (inTable) referencesInTables.add(id);
+  };
+
+  // Definition texts per id, in first-appearance order of the id.
+  const defTextsById = new Map<string, string[]>();
+
+  // Same lexer the importer uses, so the analysis matches exactly what import
+  // keeps/strips (#166): fenced lines are inert, definition lines are pulled.
+  for (const tok of lexFootnoteLines(markdown)) {
+    if (tok.inFence) continue;
+    if (tok.definition) {
+      const { id, text } = tok.definition;
+      const arr = defTextsById.get(id);
+      if (arr) arr.push(text);
+      else defTextsById.set(id, [text]);
+      // A definition's TEXT can itself reference another footnote (`[^a]: see
+      // [^b]`); count those so such a `[^b]` is not falsely reported dangling.
+      forEachFootnoteReference(text, (rid) => addRef(rid, false));
       continue;
     }
-    if (fence !== null) continue; // inside a fence: inert example text
-    if (FOOTNOTE_DEF_RE.test(line)) return true;
+    const inTable = tok.line.trimStart().startsWith("|");
+    forEachFootnoteReference(tok.line, (id) => addRef(id, inTable));
   }
-  return false;
+
+  const danglingReferences = refIds.filter((id) => !defTextsById.has(id));
+  const duplicateDefinitions: string[] = [];
+  const emptyDefinitions: string[] = [];
+  for (const [id, texts] of defTextsById) {
+    if (texts.length >= 2) duplicateDefinitions.push(id);
+    // First-wins: the kept definition is the first one; flag it if it is blank.
+    if ((texts[0] ?? "").trim().length === 0) emptyDefinitions.push(id);
+  }
+  const tableRefs = [...referencesInTables];
+
+  const warnings: string[] = [];
+  const list = (ids: string[]) => ids.map((id) => `[^${id}]`).join(", ");
+  if (danglingReferences.length > 0) {
+    warnings.push(
+      `Footnote reference(s) with no matching definition: ${list(danglingReferences)} (each will render as an empty footnote in the editor).`,
+    );
+  }
+  if (emptyDefinitions.length > 0) {
+    warnings.push(
+      `Footnote definition(s) with empty text: ${list(emptyDefinitions)}.`,
+    );
+  }
+  if (duplicateDefinitions.length > 0) {
+    warnings.push(
+      `Footnote id(s) defined more than once (only the first definition was kept): ${list(duplicateDefinitions)}.`,
+    );
+  }
+  if (tableRefs.length > 0) {
+    warnings.push(
+      `Footnote marker(s) inside a table row (footnotes in table cells may not render as expected): ${list(tableRefs)}.`,
+    );
+  }
+
+  return {
+    danglingReferences,
+    emptyDefinitions,
+    duplicateDefinitions,
+    referencesInTables: tableRefs,
+    warnings,
+  };
 }
 
 /**
  * The optional `footnoteWarnings` field for a page-write tool result: present
- * (with the single advisory) only when `markdown` uses legacy reference-style
- * footnote syntax, omitted otherwise. One helper so all three call sites
- * (create/update/import) attach the field identically. Spread into the result:
- * `{ ...result, ...footnoteWarningsField(text) }`.
+ * (with the warning lines) only when `markdown` has footnote problems, omitted
+ * otherwise. One helper so all three call sites (create/update/import) attach the
+ * field identically. Spread into the result: `{ ...result, ...footnoteWarningsField(text) }`.
  */
 export function footnoteWarningsField(markdown: string): {
   footnoteWarnings?: string[];
 } {
-  return hasLegacyFootnoteDefinition(markdown)
-    ? { footnoteWarnings: [LEGACY_FOOTNOTE_WARNING] }
-    : {};
+  const { warnings } = analyzeFootnotes(markdown);
+  return warnings.length > 0 ? { footnoteWarnings: warnings } : {};
 }

packages/mcp/src/lib/footnote-authoring.ts

@@ -0,0 +1,91 @@
+/**
+ * Inline-authoring helpers for footnotes (MCP).
+ *
+ * These build/identify footnote DEFINITION nodes for the author-inline tool
+ * (`insertInlineFootnote` in transforms.ts): a content key to de-duplicate notes
+ * by text, a definition-node factory, and a fresh uuidv7-style id generator.
+ *
+ * Split out of `footnote-canonicalize.ts` so that module stays a pure MIRROR of
+ * the editor-ext canonicalizer (compositionally symmetric to the editor-ext
+ * copy, which keeps its authoring helpers in `footnote-util.ts`). The pure
+ * canonicalizer has no dependency on these.
+ */
+
+const FOOTNOTE_DEFINITION_NAME = "footnoteDefinition";
+
+function cloneJson<T>(v: T): T {
+  if (typeof structuredClone === "function") return structuredClone(v);
+  return JSON.parse(JSON.stringify(v)) as T;
+}
+
+/**
+ * Normalized content key for de-duplicating footnote DEFINITIONS by their text.
+ *
+ * Two definitions with the same key are the SAME footnote — so the inline
+ * authoring tool reuses one id (one number, one definition, several references)
+ * instead of minting a second definition. Key = plaintext (whitespace-collapsed,
+ * trimmed) PLUS a signature of the inline mark types in order, so two notes that
+ * read the same but differ in formatting (one bold, one plain) are NOT merged.
+ * Conservative: only an exact match merges.
+ */
+export function footnoteContentKey(defNode: any): string {
+  const parts: string[] = [];
+  const visit = (n: any): void => {
+    if (!n || typeof n !== "object") return;
+    if (n.type === "text" && typeof n.text === "string") {
+      const marks = Array.isArray(n.marks)
+        ? n.marks.map((m: any) => m?.type).filter(Boolean).sort().join(",")
+        : "";
+      parts.push(`${n.text}${marks}`);
+    }
+    if (Array.isArray(n.content)) for (const c of n.content) visit(c);
+  };
+  visit(defNode);
+  // Collapse the assembled text's whitespace and trim, keeping the mark
+  // signature attached so formatting differences still distinguish notes.
+  return parts
+    .join("")
+    .replace(/[ \t\r\n]+/g, " ")
+    .trim();
+}
+
+/**
+ * Build a footnoteDefinition node from inline ProseMirror nodes, keyed by id.
+ */
+export function makeFootnoteDefinition(id: string, inlineNodes: any[]): any {
+  const content = Array.isArray(inlineNodes) ? cloneJson(inlineNodes) : [];
+  return {
+    type: FOOTNOTE_DEFINITION_NAME,
+    attrs: { id },
+    content: [{ type: "paragraph", content }],
+  };
+}
+
+/**
+ * Generate a uuidv7-style id (time-ordered), matching editor-ext's
+ * `generateFootnoteId`. Used for a genuinely-new inline footnote id.
+ */
+export function generateFootnoteId(): string {
+  const now = Date.now();
+  const timeHex = now.toString(16).padStart(12, "0");
+  const rand = (length: number) => {
+    let s = "";
+    for (let i = 0; i < length; i++)
+      s += Math.floor(Math.random() * 16).toString(16);
+    return s;
+  };
+  const versioned = "7" + rand(3);
+  const variantNibble = (8 + Math.floor(Math.random() * 4)).toString(16);
+  const variant = variantNibble + rand(3);
+  return (
+    timeHex.slice(0, 8) +
+    "-" +
+    timeHex.slice(8, 12) +
+    "-" +
+    versioned +
+    "-" +
+    variant +
+    "-" +
+    rand(12)
+  );
+}

packages/mcp/src/lib/footnote-canonicalize.ts

@@ -4,8 +4,8 @@
  * `canonicalizeFootnotes(doc)` is a pure ProseMirror-JSON port of the editor's
  * `footnoteSyncPlugin` end-state, identical in behaviour to
  * `@docmost/editor-ext`'s `canonicalizeFootnotes`. It is mirrored here — rather
- * than imported from editor-ext — for the SAME reason the `docmost-schema.ts`
- * nodes are mirrored: the MCP package is deliberately
+ * than imported from editor-ext — for the SAME reason `footnote-lex.ts` and the
+ * `docmost-schema.ts` nodes are mirrored: the MCP package is deliberately
  * decoupled from the browser/React-heavy editor barrel and operates on plain
  * JSON. The editor-ext copy owns the golden test against the live plugin; this
  * copy must stay behaviourally identical (a SHARED golden corpus, exercised by
@@ -13,8 +13,8 @@
  *
  * This module is the pure MIRROR only. The inline-authoring helpers
  * (`footnoteContentKey`, `makeFootnoteDefinition`, `generateFootnoteId`) used by
- * `insertInlineFootnote` live in `@docmost/prosemirror-markdown` (next to the
- * importer's `assembleFootnotes`, #414), so this file stays a pure mirror.
+ * `insertInlineFootnote` live in the sibling `footnote-authoring.ts`, so this
+ * file is compositionally symmetric to the editor-ext copy.
  *
  * Why it exists: every NON-editor write path (markdown import, update_page_json,
  * docmost_transform, insert_footnote) builds ProseMirror JSON directly, so the

packages/mcp/src/lib/footnote-lex.ts

@@ -0,0 +1,73 @@
+/**
+ * Shared, fence-aware line lexer for legacy footnote markdown (MCP-internal).
+ *
+ * Since #293 STEP 5 the markdown -> ProseMirror IMPORT path lives in the shared
+ * `@docmost/prosemirror-markdown` package (inline `^[body]` footnotes), so this
+ * lexer no longer backs an mcp importer. It now backs ONLY the import-time
+ * diagnostics (`analyzeFootnotes` in footnote-analyze.ts), which still scan the
+ * raw markdown for legacy reference-style `[^id]:` definition lines and surface
+ * advisory warnings (duplicate/orphan definitions) about content that is now
+ * inert on import. Fence-awareness (a `[^id]:` line inside a ``` / ~~~ block is
+ * NOT a definition) is the property the analyzer relies on.
+ *
+ * NOTE: this is deliberately NOT shared with editor-ext's
+ * `extractFootnoteDefinitions` — that lives in a different package and the
+ * decoupling between the editor and the MCP mirror is intentional.
+ */
+
+/** A footnote DEFINITION line: `[^id]: text` (id + text captured). */
+export const FOOTNOTE_DEF_RE = /^\[\^([^\]\s]+)\]:[ \t]*(.*)$/;
+/** Every footnote REFERENCE `[^id]` in a line (global; id captured). */
+export const FOOTNOTE_REF_RE_G = /\[\^([^\]\s]+)\]/g;
+/** Opening/closing code fence marker (``` or ~~~). */
+const FENCE_RE = /^(\s*)(`{3,}|~{3,})/;
+
+export interface FootnoteLine {
+  /** The raw line, verbatim. */
+  line: string;
+  /**
+   * True for a code-fence marker line AND every line inside a fence — footnote
+   * syntax on such lines is inert (example text, not real markup). The importer
+   * keeps these in the body; the analyzer skips them.
+   */
+  inFence: boolean;
+  /** The parsed definition, when this is a `[^id]: text` line OUTSIDE any fence. */
+  definition: { id: string; text: string } | null;
+}
+
+/** Classify every line of `markdown`, tracking fenced-code state. Pure. */
+export function lexFootnoteLines(markdown: string): FootnoteLine[] {
+  const out: FootnoteLine[] = [];
+  let fence: string | null = null;
+  for (const line of markdown.split("\n")) {
+    const fenceMatch = FENCE_RE.exec(line);
+    if (fenceMatch) {
+      const marker = fenceMatch[2][0];
+      if (fence === null) fence = marker; // opening fence
+      else if (marker === fence) fence = null; // matching closing fence
+      out.push({ line, inFence: true, definition: null });
+      continue;
+    }
+    if (fence !== null) {
+      out.push({ line, inFence: true, definition: null });
+      continue;
+    }
+    const m = FOOTNOTE_DEF_RE.exec(line);
+    out.push({
+      line,
+      inFence: false,
+      definition: m ? { id: m[1], text: m[2] } : null,
+    });
+  }
+  return out;
+}
+
+/** Scan a line for every `[^id]` reference, invoking `onRef(id)` for each. */
+export function forEachFootnoteReference(
+  line: string,
+  onRef: (id: string) => void,
+): void {
+  FOOTNOTE_REF_RE_G.lastIndex = 0;
+  let m: RegExpExecArray | null;
+  while ((m = FOOTNOTE_REF_RE_G.exec(line)) !== null) onRef(m[1]);
+}

packages/mcp/src/lib/node-ops.ts

@@ -0,0 +1,963 @@
+/**
+ * Pure, network-free helpers for manipulating a ProseMirror/TipTap document
+ * tree by node id.
+ *
+ * A ProseMirror node here is a plain JSON object of the shape produced by
+ * Docmost: `{ type, attrs?, content?, text?, marks? }`. Children live in the
+ * `content` array; a node carries a stable id in `attrs.id`. Callouts and
+ * table cells hold their children in `content` just like any other block, so a
+ * single recursive walk reaches them all.
+ *
+ * Every exported function operates on a DEEP CLONE of the input document and
+ * returns the new document. The input doc and any `newNode`/`node` argument are
+ * never mutated. All functions are defensively null-safe: missing/!Array
+ * `content`, non-object nodes, and absent `attrs` are tolerated.
+ */
+
+import { stripInlineMarkdown } from "./text-normalize.js";
+
+/** Deep-clone a JSON-serializable value without mutating the original. */
+function clone<T>(value: T): T {
+  if (typeof structuredClone === "function") {
+    return structuredClone(value);
+  }
+  // Fallback for environments without structuredClone.
+  return JSON.parse(JSON.stringify(value)) as T;
+}
+
+/** True if `value` is a non-null object (and not an array). */
+function isObject(value: any): value is Record<string, any> {
+  return value != null && typeof value === "object" && !Array.isArray(value);
+}
+
+/** True if `node` carries the given id in `node.attrs.id`. */
+function matchesId(node: any, nodeId: string): boolean {
+  return isObject(node) && isObject(node.attrs) && node.attrs.id === nodeId;
+}
+
+/**
+ * Recursively concatenate all text contained in a node.
+ *
+ * Text nodes contribute their `text` string; container nodes contribute the
+ * joined `blockPlainText` of their `content` children. Returns "" for nullish
+ * or non-object inputs.
+ */
+export function blockPlainText(node: any): string {
+  if (!isObject(node)) return "";
+  let out = "";
+  if (typeof node.text === "string") {
+    out += node.text;
+  }
+  if (Array.isArray(node.content)) {
+    for (const child of node.content) {
+      out += blockPlainText(child);
+    }
+  }
+  return out;
+}
+
+/** Truncate `text` to at most `n` chars, appending an ellipsis when cut. */
+function truncate(text: string, n: number): string {
+  return text.length > n ? text.slice(0, n) + "…" : text;
+}
+
+/** One compact outline entry for a single top-level block. */
+export interface OutlineEntry {
+  index: number;
+  type: string | undefined;
+  id: string | null;
+  firstText: string;
+  /** Present for headings only. */
+  level?: number | null;
+  /** Present for tables only. */
+  rows?: number;
+  cols?: number;
+  header?: string[];
+  /** Present for list blocks only (bulletList/orderedList/taskList). */
+  items?: number;
+}
+
+/**
+ * Build a COMPACT outline of the TOP-LEVEL blocks of `doc` (the entries in
+ * `doc.content`). Deliberately does NOT recurse into paragraphs, list items, or
+ * table cells — compactness is the point; use `getNodeByRef` to drill into a
+ * specific block.
+ *
+ * Each entry carries `{ index, type, id, firstText }`, plus type-specific
+ * extras: headings add `level`; tables add `rows`/`cols` and the first row's
+ * cell texts as `header`; list blocks (types ending in "List") add `items`.
+ * `firstText` is the block's plain text truncated to 100 chars. Null-safe:
+ * a missing or non-object doc/content yields `[]`.
+ */
+export function buildOutline(doc: any): OutlineEntry[] {
+  if (!isObject(doc) || !Array.isArray(doc.content)) return [];
+
+  const out: OutlineEntry[] = [];
+  for (let i = 0; i < doc.content.length; i++) {
+    const block = doc.content[i];
+    const type = isObject(block) ? block.type : undefined;
+    const entry: OutlineEntry = {
+      index: i,
+      type,
+      id:
+        isObject(block) && isObject(block.attrs)
+          ? (block.attrs.id ?? null)
+          : null,
+      firstText: truncate(blockPlainText(block), 100),
+    };
+
+    if (type === "heading") {
+      entry.level = isObject(block.attrs) ? (block.attrs.level ?? null) : null;
+    } else if (type === "table") {
+      const headerRow = block.content?.[0]?.content ?? [];
+      entry.rows = block.content?.length ?? 0;
+      entry.cols = block.content?.[0]?.content?.length ?? 0;
+      entry.header = headerRow.map((cell: any) =>
+        truncate(blockPlainText(cell), 40),
+      );
+    } else if (typeof type === "string" && type.endsWith("List")) {
+      entry.items = block.content?.length ?? 0;
+    }
+
+    out.push(entry);
+  }
+  return out;
+}
+
+/**
+ * Resolve a single node by reference and return `{ node, path, type }`, or
+ * `null` when nothing matches.
+ *
+ * - `ref` of the form `#<n>` (e.g. `#2`) selects the TOP-LEVEL block at index
+ *   `n` in `doc.content`. This is the only way to address table/tableRow/
+ *   tableCell nodes, which carry no `attrs.id`.
+ * - Otherwise `ref` is treated as a block id: the FIRST node anywhere in the
+ *   tree with `attrs.id === ref` is returned.
+ *
+ * `path` is the array of child indices from the doc root down to the node
+ * (so a top-level block is `[index]`). The returned `node` is a DEEP CLONE,
+ * so callers can mutate it without touching the input doc. Null-safe.
+ */
+export function getNodeByRef(
+  doc: any,
+  ref: string,
+): { node: any; path: number[]; type: string | undefined } | null {
+  if (!isObject(doc)) return null;
+
+  // "#<n>": index into the top-level content array.
+  const indexMatch = typeof ref === "string" ? ref.match(/^#(\d+)$/) : null;
+  if (indexMatch) {
+    const index = Number(indexMatch[1]);
+    const block = Array.isArray(doc.content) ? doc.content[index] : undefined;
+    if (!isObject(block)) return null;
+    return { node: clone(block), path: [index], type: block.type };
+  }
+
+  // Otherwise: depth-first search for the first node with attrs.id === ref.
+  const search = (
+    node: any,
+    trail: number[],
+  ): { node: any; path: number[]; type: string } | null => {
+    if (!isObject(node)) return null;
+    if (Array.isArray(node.content)) {
+      for (let i = 0; i < node.content.length; i++) {
+        const child = node.content[i];
+        const path = [...trail, i];
+        if (matchesId(child, ref)) {
+          return { node: clone(child), path, type: child.type };
+        }
+        const hit = search(child, path);
+        if (hit != null) return hit;
+      }
+    }
+    return null;
+  };
+
+  return search(doc, []);
+}
+
+/**
+ * Replace EVERY node whose `attrs.id === nodeId` with a deep clone of
+ * `newNode`, anywhere in the tree (including inside callouts and table cells).
+ *
+ * Operates on a clone of `doc`; returns `{ doc, replaced }` where `replaced`
+ * is the number of nodes substituted. A fresh clone of `newNode` is used for
+ * each match so they do not share references.
+ */
+export function replaceNodeById(
+  doc: any,
+  nodeId: string,
+  newNode: any,
+): { doc: any; replaced: number } {
+  const out = clone(doc);
+  let replaced = 0;
+
+  // Walk a content array, replacing direct matches and recursing into the
+  // (possibly new) children of non-matching nodes.
+  const walkContent = (content: any[]): void => {
+    for (let i = 0; i < content.length; i++) {
+      const child = content[i];
+      if (matchesId(child, nodeId)) {
+        content[i] = clone(newNode);
+        replaced++;
+        // Do not recurse into a freshly substituted node.
+        continue;
+      }
+      if (isObject(child) && Array.isArray(child.content)) {
+        walkContent(child.content);
+      }
+    }
+  };
+
+  if (isObject(out) && Array.isArray(out.content)) {
+    walkContent(out.content);
+  }
+  return { doc: out, replaced };
+}
+
+/**
+ * Remove EVERY node whose `attrs.id === nodeId` from its parent `content`
+ * array, anywhere in the tree (recursive, including callouts and tables).
+ *
+ * Operates on a clone of `doc`; returns `{ doc, deleted }` where `deleted` is
+ * the number of nodes removed.
+ */
+export function deleteNodeById(
+  doc: any,
+  nodeId: string,
+): { doc: any; deleted: number } {
+  const out = clone(doc);
+  let deleted = 0;
+
+  // Filter a content array in place, dropping matches and recursing into the
+  // surviving children.
+  const walkContent = (content: any[]): any[] => {
+    const kept: any[] = [];
+    for (const child of content) {
+      if (matchesId(child, nodeId)) {
+        deleted++;
+        continue;
+      }
+      if (isObject(child) && Array.isArray(child.content)) {
+        child.content = walkContent(child.content);
+      }
+      kept.push(child);
+    }
+    return kept;
+  };
+
+  if (isObject(out) && Array.isArray(out.content)) {
+    out.content = walkContent(out.content);
+  }
+  return { doc: out, deleted };
+}
+
+/**
+ * Throw a clear, model-actionable error when a node-id write op did NOT match
+ * exactly one node (#159). `count === 0` -> "no node found"; `count > 1` ->
+ * "ambiguous, refused" — Docmost duplicates block ids on copy/paste, so a write
+ * by id could clobber/remove EVERY duplicate. The caller skips the write for any
+ * `count !== 1` (the transform returns null), so this only REPORTS; nothing was
+ * changed. No-op for the unambiguous single-match case.
+ */
+export function assertUnambiguousMatch(
+  op: "patch_node" | "delete_node",
+  verb: "replace" | "delete",
+  count: number,
+  nodeId: string,
+  pageId: string,
+): void {
+  if (count === 0) {
+    throw new Error(
+      `${op}: no node with id "${nodeId}" found on page ${pageId}`,
+    );
+  }
+  if (count > 1) {
+    throw new Error(
+      `${op}: id "${nodeId}" is ambiguous — ${count} nodes on page ${pageId} share it (block ids are duplicated on copy/paste). Refusing to ${verb} all of them; nothing was changed. Re-target with a more specific anchor.`,
+    );
+  }
+}
+
+/**
+ * Deep-clone `doc` and strip every node/mark attribute whose value is strictly
+ * `undefined`, so the result is safe to hand to Yjs (which throws an opaque
+ * "Unexpected content type" when asked to store an `undefined` attribute value).
+ *
+ * Only `undefined` keys are removed; `null`, `false`, `0`, and `""` are all
+ * legitimate JSON-storable values and are preserved. Operates on a clone and
+ * returns it; the input is never mutated. Defensively null-safe like the rest
+ * of the file.
+ */
+export function sanitizeForYjs(doc: any): any {
+  const out = clone(doc);
+
+  // Drop every key whose value is strictly `undefined` from an attrs object.
+  const stripUndefined = (attrs: any): void => {
+    if (!isObject(attrs)) return;
+    for (const key of Object.keys(attrs)) {
+      if (attrs[key] === undefined) {
+        delete attrs[key];
+      }
+    }
+  };
+
+  const walk = (node: any): void => {
+    if (!isObject(node)) return;
+    stripUndefined(node.attrs);
+    if (Array.isArray(node.marks)) {
+      for (const mark of node.marks) {
+        if (isObject(mark)) stripUndefined(mark.attrs);
+      }
+    }
+    if (Array.isArray(node.content)) {
+      for (const child of node.content) {
+        walk(child);
+      }
+    }
+  };
+
+  walk(out);
+  return out;
+}
+
+/**
+ * Diagnostics helper: walk the tree and return a human-readable path string for
+ * the FIRST attribute value (in any `node.attrs` or `mark.attrs`) that Yjs
+ * cannot store — i.e. `undefined`, a `function`, a `symbol`, or a `bigint`
+ * (e.g. `content[3].content[0].attrs.indent (undefined)`). Returns `null` when
+ * every attribute is storable. Null-safe.
+ */
+export function findUnstorableAttr(doc: any): string | null {
+  const isUnstorable = (value: any): string | null => {
+    if (value === undefined) return "undefined";
+    const t = typeof value;
+    if (t === "function") return "function";
+    if (t === "symbol") return "symbol";
+    if (t === "bigint") return "bigint";
+    return null;
+  };
+
+  // Check an attrs object; return the offending sub-path or null.
+  const checkAttrs = (attrs: any, basePath: string): string | null => {
+    if (!isObject(attrs)) return null;
+    for (const key of Object.keys(attrs)) {
+      const kind = isUnstorable(attrs[key]);
+      if (kind != null) return `${basePath}.${key} (${kind})`;
+    }
+    return null;
+  };
+
+  const walk = (node: any, path: string): string | null => {
+    if (!isObject(node)) return null;
+    const attrHit = checkAttrs(node.attrs, `${path}.attrs`);
+    if (attrHit != null) return attrHit;
+    if (Array.isArray(node.marks)) {
+      for (let i = 0; i < node.marks.length; i++) {
+        const markHit = checkAttrs(
+          node.marks[i]?.attrs,
+          `${path}.marks[${i}].attrs`,
+        );
+        if (markHit != null) return markHit;
+      }
+    }
+    if (Array.isArray(node.content)) {
+      for (let i = 0; i < node.content.length; i++) {
+        const childHit = walk(node.content[i], `${path}.content[${i}]`);
+        if (childHit != null) return childHit;
+      }
+    }
+    return null;
+  };
+
+  // The root doc node carries no useful index, so start the path at "doc".
+  if (!isObject(doc)) return null;
+  const attrHit = checkAttrs(doc.attrs, "attrs");
+  if (attrHit != null) return attrHit;
+  if (Array.isArray(doc.content)) {
+    for (let i = 0; i < doc.content.length; i++) {
+      const childHit = walk(doc.content[i], `content[${i}]`);
+      if (childHit != null) return childHit;
+    }
+  }
+  return null;
+}
+
+/**
+ * Table structural node types and the container each must live directly inside.
+ * Used by `insertNodeRelative` to splice rows/cells into the correct ancestor
+ * rather than blindly into the anchor's direct parent (which would corrupt the
+ * table's nesting).
+ */
+const STRUCTURAL_TYPES = new Set(["tableRow", "tableCell", "tableHeader"]);
+const REQUIRED_CONTAINER: Record<string, string> = {
+  tableRow: "table",
+  tableCell: "tableRow",
+  tableHeader: "tableRow",
+};
+
+/**
+ * Find the index of the first TOP-LEVEL block whose plain text includes the
+ * anchor, with a markdown-stripping FALLBACK. Returns -1 when none matches.
+ *
+ * Two passes preserve "exact wins globally":
+ *  - Pass 1: first block containing the verbatim `anchorText`.
+ *  - Pass 2 (only if pass 1 found nothing): first block containing the
+ *    markdown-stripped anchor, when stripping actually changed it.
+ */
+function findAnchorTextIndex(content: any[], anchorText: string): number {
+  if (!Array.isArray(content)) return -1;
+  // Pass 1: exact.
+  for (let i = 0; i < content.length; i++) {
+    if (blockPlainText(content[i]).includes(anchorText)) return i;
+  }
+  // Pass 2: markdown-stripped fallback.
+  const a = stripInlineMarkdown(anchorText);
+  if (a !== anchorText && a.length > 0) {
+    for (let i = 0; i < content.length; i++) {
+      if (blockPlainText(content[i]).includes(a)) return i;
+    }
+  }
+  return -1;
+}
+
+/**
+ * Locate an anchor and return its ancestor chain (from `doc` down to and
+ * including the matched node). Each chain entry is `{ node, index }` where
+ * `index` is the node's position inside its parent's `content` array (the root
+ * doc has index -1). Returns `null` when the anchor cannot be resolved.
+ */
+function findAnchorChain(
+  doc: any,
+  opts: InsertOptions,
+): { node: any; index: number }[] | null {
+  if (!isObject(doc)) return null;
+
+  // DFS by id anywhere in the tree, accumulating the path.
+  if (opts.anchorNodeId != null) {
+    const targetId = opts.anchorNodeId;
+    const search = (
+      node: any,
+      index: number,
+      trail: { node: any; index: number }[],
+    ): { node: any; index: number }[] | null => {
+      if (!isObject(node)) return null;
+      const here = [...trail, { node, index }];
+      if (matchesId(node, targetId)) return here;
+      if (Array.isArray(node.content)) {
+        for (let i = 0; i < node.content.length; i++) {
+          const hit = search(node.content[i], i, here);
+          if (hit != null) return hit;
+        }
+      }
+      return null;
+    };
+    return search(doc, -1, []);
+  }
+
+  // By text: only top-level blocks are scanned (same rule as the JSON path).
+  // Exact match wins; a markdown-stripped fallback is tried only on a miss.
+  if (opts.anchorText != null && Array.isArray(doc.content)) {
+    const i = findAnchorTextIndex(doc.content, opts.anchorText);
+    if (i !== -1) {
+      return [
+        { node: doc, index: -1 },
+        { node: doc.content[i], index: i },
+      ];
+    }
+  }
+
+  return null;
+}
+
+/** Options controlling where `insertNodeRelative` places the new node. */
+export interface InsertOptions {
+  position: "before" | "after" | "append";
+  /** Resolve the anchor by node id anywhere in the tree (preferred). */
+  anchorNodeId?: string;
+  /** Fallback: first TOP-LEVEL block whose plain text includes this string. */
+  anchorText?: string;
+}
+
+/**
+ * Insert a deep clone of `node` relative to an anchor.
+ *
+ * - position "append": push the node onto the top-level `doc.content`.
+ * - position "before"/"after": locate the anchor and splice the node into the
+ *   anchor's parent `content` array immediately before / after it.
+ *
+ * Anchor resolution for before/after:
+ *   - if `anchorNodeId` is given, find the node with `attrs.id === anchorNodeId`
+ *     anywhere in the tree (recursive);
+ *   - otherwise, if `anchorText` is given, scan only TOP-LEVEL `doc.content`
+ *     blocks and pick the first whose `blockPlainText` includes `anchorText`.
+ *
+ * Operates on a clone of `doc`; returns `{ doc, inserted }`. `inserted` is
+ * false when the anchor could not be resolved (the doc is returned unchanged
+ * apart from being cloned).
+ */
+export function insertNodeRelative(
+  doc: any,
+  node: any,
+  opts: InsertOptions,
+): { doc: any; inserted: boolean } {
+  const out = clone(doc);
+  const fresh = clone(node);
+
+  // Defensive: stay null-safe like the other exports — a missing opts means
+  // there is nothing actionable to do.
+  if (!isObject(opts)) return { doc: out, inserted: false };
+
+  const isStructural = isObject(node) && STRUCTURAL_TYPES.has(node.type);
+
+  // "append": top-level push.
+  if (opts.position === "append") {
+    // Structural table nodes (tableRow/tableCell/tableHeader) cannot live at the
+    // top level — appending one would produce invalid nesting.
+    if (isStructural) {
+      throw new Error(
+        `insert_node: cannot append a ${node.type} at the top level; use ` +
+          `position before/after with an anchor inside the target table`,
+      );
+    }
+    if (isObject(out)) {
+      if (!Array.isArray(out.content)) out.content = [];
+      out.content.push(fresh);
+      return { doc: out, inserted: true };
+    }
+    return { doc: out, inserted: false };
+  }
+
+  const offset = opts.position === "after" ? 1 : 0;
+
+  // Structural insert (before/after a tableRow/tableCell/tableHeader): splice
+  // into the nearest enclosing table/tableRow rather than the anchor's direct
+  // parent, so the row/cell lands at the correct level of the table.
+  if (isStructural) {
+    const containerType = REQUIRED_CONTAINER[node.type];
+    const chain = findAnchorChain(out, opts);
+    // Anchor not resolved at all — keep the existing "anchor not found" path.
+    if (chain == null) return { doc: out, inserted: false };
+
+    // Find the DEEPEST ancestor (including the anchor itself) of the required
+    // container type.
+    let containerIdx = -1;
+    for (let i = chain.length - 1; i >= 0; i--) {
+      if (isObject(chain[i].node) && chain[i].node.type === containerType) {
+        containerIdx = i;
+        break;
+      }
+    }
+
+    if (containerIdx === -1) {
+      throw new Error(
+        `insert_node: cannot insert a ${node.type} here — the anchor is not ` +
+          `inside a ${containerType}. Anchor on a cell's text or a block id ` +
+          `that lives inside the target table.`,
+      );
+    }
+
+    const container = chain[containerIdx].node;
+    if (!Array.isArray(container.content)) container.content = [];
+
+    if (containerIdx === chain.length - 1) {
+      // The matched container IS the anchor node itself (e.g. anchorText
+      // resolved to the table block): append/prepend within it.
+      const at = opts.position === "after" ? container.content.length : 0;
+      container.content.splice(at, 0, fresh);
+    } else {
+      // The immediate child on the path leading to the anchor is the row/cell
+      // to splice next to.
+      const enclosingChildIndex = chain[containerIdx + 1].index;
+      container.content.splice(enclosingChildIndex + offset, 0, fresh);
+    }
+    return { doc: out, inserted: true };
+  }
+
+  // Resolve by id anywhere in the tree: splice into the parent content array.
+  if (opts.anchorNodeId != null) {
+    let inserted = false;
+    const walkContent = (content: any[]): void => {
+      for (let i = 0; i < content.length; i++) {
+        const child = content[i];
+        if (matchesId(child, opts.anchorNodeId as string)) {
+          content.splice(i + offset, 0, fresh);
+          inserted = true;
+          return;
+        }
+        if (isObject(child) && Array.isArray(child.content)) {
+          walkContent(child.content);
+          if (inserted) return;
+        }
+      }
+    };
+    if (isObject(out) && Array.isArray(out.content)) {
+      walkContent(out.content);
+    }
+    return { doc: out, inserted };
+  }
+
+  // Resolve by text: only top-level doc.content blocks are scanned. Exact
+  // match wins; a markdown-stripped fallback is tried only on a miss.
+  if (opts.anchorText != null && isObject(out) && Array.isArray(out.content)) {
+    const i = findAnchorTextIndex(out.content, opts.anchorText);
+    if (i !== -1) {
+      out.content.splice(i + offset, 0, fresh);
+      return { doc: out, inserted: true };
+    }
+  }
+
+  return { doc: out, inserted: false };
+}
+
+// ===========================================================================
+// Table editing helpers
+//
+// A Docmost table is a ProseMirror subtree with NO ids on the structural nodes:
+//   table   -> { type:"table",     content:[tableRow...] }
+//   row     -> { type:"tableRow",  content:[tableCell|tableHeader...] }
+//   cell    -> { type:"tableCell"|"tableHeader", attrs:{colspan,rowspan,colwidth},
+//                content:[paragraph...] }
+//   para    -> { type:"paragraph", attrs:{id,indent}, content:[textNode...] }
+// Only paragraphs/headings carry an `attrs.id`, so a cell is addressed via the
+// id of the paragraph inside it. The helpers below all operate on a DEEP CLONE
+// of the input doc (via `clone`) and never mutate their inputs.
+// ===========================================================================
+
+/**
+ * Collect EVERY `attrs.id` present anywhere in `node` into `used`. Used to seed
+ * `makeFreshId` so generated paragraph ids never collide with existing ones.
+ */
+function collectIds(node: any, used: Set<string>): void {
+  if (!isObject(node)) return;
+  if (isObject(node.attrs) && typeof node.attrs.id === "string") {
+    used.add(node.attrs.id);
+  }
+  if (Array.isArray(node.content)) {
+    for (const child of node.content) collectIds(child, used);
+  }
+}
+
+/**
+ * Fresh-id generator: returns a random Docmost-style id (12 chars from
+ * lowercase `a-z0-9`) that is not already in `used`, and records it. On the
+ * rare collision the id is regenerated. Callers rely on uniqueness, not on the
+ * exact string, so randomness is fine — and unlike a module-local counter it
+ * needs no reset and cannot become predictable across calls.
+ */
+function makeFreshId(used: Set<string>): string {
+  const alphabet = "abcdefghijklmnopqrstuvwxyz0123456789";
+  let id: string;
+  do {
+    id = "";
+    for (let i = 0; i < 12; i++) {
+      id += alphabet[Math.floor(Math.random() * alphabet.length)];
+    }
+  } while (used.has(id) || id === "");
+  used.add(id);
+  return id;
+}
+
+/**
+ * Resolve a table reference against an ALREADY-CLONED doc and return the LIVE
+ * table node (a reference inside `rootClone`, so the caller may mutate it) plus
+ * its index path. Returns null when no table matches.
+ *
+ * - `#<n>`: the top-level block at index `n`, only if its `type === "table"`.
+ * - otherwise: DFS for the node with `attrs.id === tableRef`, then walk UP its
+ *   ancestor chain to the nearest `type === "table"` ancestor.
+ */
+function locateTable(
+  rootClone: any,
+  tableRef: string,
+): { table: any; path: number[] } | null {
+  if (!isObject(rootClone)) return null;
+
+  // "#<n>": index into the top-level content array; must be a table.
+  const indexMatch =
+    typeof tableRef === "string" ? tableRef.match(/^#(\d+)$/) : null;
+  if (indexMatch) {
+    const index = Number(indexMatch[1]);
+    const block = Array.isArray(rootClone.content)
+      ? rootClone.content[index]
+      : undefined;
+    if (isObject(block) && block.type === "table") {
+      return { table: block, path: [index] };
+    }
+    return null;
+  }
+
+  // Otherwise: DFS for attrs.id === tableRef, tracking the ancestor chain, then
+  // climb to the nearest enclosing table.
+  const search = (
+    node: any,
+    trail: { node: any; index: number }[],
+  ): { table: any; path: number[] } | null => {
+    if (!isObject(node)) return null;
+    if (Array.isArray(node.content)) {
+      for (let i = 0; i < node.content.length; i++) {
+        const child = node.content[i];
+        const here = [...trail, { node: child, index: i }];
+        if (matchesId(child, tableRef)) {
+          // Walk UP to the nearest table ancestor (including the match itself).
+          for (let j = here.length - 1; j >= 0; j--) {
+            if (isObject(here[j].node) && here[j].node.type === "table") {
+              return {
+                table: here[j].node,
+                path: here.slice(0, j + 1).map((e) => e.index),
+              };
+            }
+          }
+          return null; // id found but no enclosing table
+        }
+        const hit = search(child, here);
+        if (hit != null) return hit;
+      }
+    }
+    return null;
+  };
+
+  return search(rootClone, []);
+}
+
+/** Build the plain-text → single-paragraph cell content used by all writers. */
+function makeCellParagraph(id: string, text: string): any {
+  return {
+    type: "paragraph",
+    attrs: { id, indent: 0 },
+    // Empty string → a paragraph with an empty content array.
+    content: text ? [{ type: "text", text }] : [],
+  };
+}
+
+/**
+ * Read a table as a matrix. Returns null when `tableRef` resolves to no table.
+ *
+ * - `rows`/`cols`: the table's row count and the column count of its FIRST row.
+ *   Tables may be ragged (rows of differing length), so `cols` reflects only
+ *   row 0; use the per-row length of `cells`/`cellIds` for each row's actual
+ *   width.
+ * - `cells`: `string[][]` of each cell's `blockPlainText`.
+ * - `cellIds`: `(string|null)[][]` of each cell's FIRST paragraph id (or null),
+ *   so callers can `patch_node` a cell for rich-formatted edits.
+ * - `path`: index path of the table within the doc.
+ */
+export function readTable(
+  doc: any,
+  tableRef: string,
+): {
+  rows: number;
+  cols: number;
+  cells: string[][];
+  cellIds: (string | null)[][];
+  path: number[];
+} | null {
+  const root = clone(doc);
+  const located = locateTable(root, tableRef);
+  if (located == null) return null;
+  const { table, path } = located;
+
+  const rowNodes = Array.isArray(table.content) ? table.content : [];
+  const rows = rowNodes.length;
+  const cols = rowNodes[0]?.content?.length ?? 0;
+
+  const cells: string[][] = [];
+  const cellIds: (string | null)[][] = [];
+  for (const rowNode of rowNodes) {
+    const cellNodes = Array.isArray(rowNode?.content) ? rowNode.content : [];
+    const rowText: string[] = [];
+    const rowIds: (string | null)[] = [];
+    for (const cellNode of cellNodes) {
+      rowText.push(blockPlainText(cellNode));
+      // The cell's first paragraph carries the id used for patch_node.
+      const firstPara = Array.isArray(cellNode?.content)
+        ? cellNode.content[0]
+        : undefined;
+      const id =
+        isObject(firstPara) && isObject(firstPara.attrs)
+          ? (firstPara.attrs.id ?? null)
+          : null;
+      rowIds.push(id);
+    }
+    cells.push(rowText);
+    cellIds.push(rowIds);
+  }
+
+  return { rows, cols, cells, cellIds, path };
+}
+
+/**
+ * Insert a row of plain-text cells into a table. Returns `{ doc, inserted }`.
+ *
+ * The row is padded to the table's column count (`cells[i] ?? ""`); supplying
+ * MORE cells than columns throws. Each new cell copies `colwidth` for its
+ * column from the header row when present, gets a fresh-id paragraph, and a
+ * `colspan:1, rowspan:1` attrs. `index` (when an integer in `[0, rows]`) splices
+ * the row there; otherwise the row is appended at the end.
+ */
+export function insertTableRow(
+  doc: any,
+  tableRef: string,
+  cells: string[],
+  index?: number,
+): { doc: any; inserted: boolean } {
+  const out = clone(doc);
+  const located = locateTable(out, tableRef);
+  if (located == null) return { doc: out, inserted: false };
+  const { table } = located;
+
+  if (!Array.isArray(table.content)) table.content = [];
+  const rows = table.content.length;
+  const headerRow = table.content[0];
+  const headerCells = Array.isArray(headerRow?.content)
+    ? headerRow.content
+    : [];
+
+  // Column count is the WIDEST existing row, so the guard below stays
+  // meaningful for ragged tables and the new row matches the table's width.
+  // Fall back to the supplied cell count only when the table has no rows.
+  let colCount = 0;
+  for (const r of table.content) {
+    if (isObject(r) && Array.isArray(r.content))
+      colCount = Math.max(colCount, r.content.length);
+  }
+  if (colCount === 0) colCount = Array.isArray(cells) ? cells.length : 0;
+
+  if (Array.isArray(cells) && cells.length > colCount) {
+    throw new Error(
+      `table_insert_row: got ${cells.length} cell(s) but the table has ${colCount} column(s)`,
+    );
+  }
+
+  // Resolve the landing index up front so the cell-type decision and the splice
+  // below agree: a valid integer in [0, rows] splices there, else we append.
+  const landingIndex =
+    typeof index === "number" &&
+    Number.isInteger(index) &&
+    index >= 0 &&
+    index <= rows
+      ? index
+      : rows;
+
+  // Seed the id generator with every id already in the doc so the new cell
+  // paragraph ids are unique within the whole document.
+  const used = new Set<string>();
+  collectIds(out, used);
+
+  const newCells: any[] = [];
+  for (let i = 0; i < colCount; i++) {
+    const text = (Array.isArray(cells) ? cells[i] : undefined) ?? "";
+    const attrs: Record<string, any> = { colspan: 1, rowspan: 1 };
+    // Copy this column's colwidth from the header row's cell when present.
+    const colwidth = headerCells[i]?.attrs?.colwidth;
+    if (colwidth !== undefined) attrs.colwidth = colwidth;
+    // A row landing at index 0 becomes the new header row, so inherit the
+    // current header cell's type per column (Docmost uses "tableHeader" there);
+    // every other position is a plain data cell.
+    const cellType =
+      landingIndex === 0 ? (headerCells[i]?.type ?? "tableCell") : "tableCell";
+    newCells.push({
+      type: cellType,
+      attrs,
+      content: [makeCellParagraph(makeFreshId(used), text)],
+    });
+  }
+
+  const newRow = { type: "tableRow", content: newCells };
+
+  // Splice at the resolved landing index (append when index was omitted/invalid).
+  table.content.splice(landingIndex, 0, newRow);
+
+  return { doc: out, inserted: true };
+}
+
+/**
+ * Delete the row at 0-based `index` from a table. Returns `{ doc, deleted }`.
+ * `deleted` is false only when the table cannot be located. Throws on an
+ * out-of-range index, and refuses to delete the table's only row.
+ */
+export function deleteTableRow(
+  doc: any,
+  tableRef: string,
+  index: number,
+): { doc: any; deleted: boolean } {
+  const out = clone(doc);
+  const located = locateTable(out, tableRef);
+  if (located == null) return { doc: out, deleted: false };
+  const { table } = located;
+
+  if (!Array.isArray(table.content)) table.content = [];
+  const rows = table.content.length;
+
+  if (!Number.isInteger(index) || index < 0 || index >= rows) {
+    throw new Error(
+      `table_delete_row: row index ${index} out of range (table has ${rows} row(s))`,
+    );
+  }
+  if (rows <= 1) {
+    throw new Error(
+      "table_delete_row: refusing to delete the only row of the table",
+    );
+  }
+
+  table.content.splice(index, 1);
+  return { doc: out, deleted: true };
+}
+
+/**
+ * Set the plain-text content of cell `[row, col]` (0-based) to `text`. Returns
+ * `{ doc, updated }`; `updated` is false only when the table cannot be located.
+ * Throws when `row`/`col` is out of range. The cell's own attrs (colspan/
+ * rowspan/colwidth) are preserved; its content becomes a single text paragraph
+ * that reuses the cell's existing first-paragraph id when present, else a fresh
+ * one.
+ */
+export function updateTableCell(
+  doc: any,
+  tableRef: string,
+  row: number,
+  col: number,
+  text: string,
+): { doc: any; updated: boolean } {
+  const out = clone(doc);
+  const located = locateTable(out, tableRef);
+  if (located == null) return { doc: out, updated: false };
+  const { table } = located;
+
+  const rowNodes = Array.isArray(table.content) ? table.content : [];
+  const rows = rowNodes.length;
+  const rowNode = rowNodes[row];
+  const cols =
+    isObject(rowNode) && Array.isArray(rowNode.content)
+      ? rowNode.content.length
+      : 0;
+
+  if (
+    !Number.isInteger(row) ||
+    row < 0 ||
+    row >= rows ||
+    !Number.isInteger(col) ||
+    col < 0 ||
+    col >= cols
+  ) {
+    throw new Error(`table_update_cell: cell [${row},${col}] out of range`);
+  }
+
+  const cellNode = rowNode.content[col];
+  // Reuse the cell's existing first-paragraph id, or mint a fresh unique one.
+  const existingPara = Array.isArray(cellNode?.content)
+    ? cellNode.content[0]
+    : undefined;
+  let id =
+    isObject(existingPara) && isObject(existingPara.attrs)
+      ? existingPara.attrs.id
+      : undefined;
+  if (typeof id !== "string" || id.length === 0) {
+    const used = new Set<string>();
+    collectIds(out, used);
+    id = makeFreshId(used);
+  }
+
+  cellNode.content = [makeCellParagraph(id, text)];
+  return { doc: out, updated: true };
+}

packages/mcp/src/lib/page-search.ts

@@ -33,7 +33,7 @@
 
 import RE2 from "re2";
 
-import { blockPlainText } from "@docmost/prosemirror-markdown";
+import { blockPlainText } from "./node-ops.js";
 
 /** An RE2 regex instance (RE2 extends `RegExp`, so it is usable as one). */
 type Re2Regex = InstanceType<typeof RE2>;

packages/mcp/src/lib/pako.d.ts

@@ -0,0 +1,61 @@
+// Minimal ambient type declaration for `pako` (no @types/pako is installed and
+// pako 2.x ships no bundled .d.ts). We only use the raw-deflate codec to read
+// draw.io's compressed `<diagram>` payload, so declare just that surface.
+declare module "pako" {
+  interface RawOptions {
+    /** When "string", the result is returned as a (binary/UTF-8) string. */
+    to?: "string";
+    /** Raw-deflate window bits; draw.io uses raw deflate (no zlib header). */
+    windowBits?: number;
+    level?: number;
+  }
+
+  /** Raw-inflate (windowBits: -15). `to:"string"` yields a string. */
+  export function inflateRaw(
+    data: Uint8Array | ArrayBuffer | number[],
+    options: RawOptions & { to: "string" },
+  ): string;
+  export function inflateRaw(
+    data: Uint8Array | ArrayBuffer | number[],
+    options?: RawOptions,
+  ): Uint8Array;
+
+  /** Raw-deflate (windowBits: -15). Used only by tests to build fixtures. */
+  export function deflateRaw(
+    data: Uint8Array | string,
+    options?: RawOptions,
+  ): Uint8Array;
+
+  interface InflateStreamOptions {
+    to?: "string";
+    windowBits?: number;
+    /** Raw deflate (no zlib header) — equivalent to windowBits: -15. */
+    raw?: boolean;
+    chunkSize?: number;
+  }
+
+  /**
+   * Streaming inflate. We use it to bound the decompressed size: `onData` is
+   * invoked per output chunk, letting us abort a decompression bomb before the
+   * full output is materialised.
+   */
+  export class Inflate {
+    constructor(options?: InflateStreamOptions);
+    onData: (chunk: string | Uint8Array) => void;
+    onEnd: (status: number) => void;
+    push(
+      data: Uint8Array | ArrayBuffer | number[] | string,
+      flushMode?: boolean | number,
+    ): boolean;
+    result: string | Uint8Array;
+    err: number;
+    msg: string;
+  }
+
+  const _default: {
+    inflateRaw: typeof inflateRaw;
+    deflateRaw: typeof deflateRaw;
+    Inflate: typeof Inflate;
+  };
+  export default _default;
+}

packages/mcp/src/lib/parse-node-arg.ts

@@ -2,11 +2,6 @@
 // instead of an object. Normalize: parse a string to an object (throwing on
 // invalid JSON), pass an object through unchanged. Shared by patch_node /
 // insert_node (and the analogous update_page_json content parsing).
-//
-// This lives in the converter package (#414) so BOTH consumers import the ONE
-// copy: `@docmost/mcp` (ESM) and the CommonJS server app. The server cannot
-// import `@docmost/mcp` directly (ESM-only, no declaration files), but it does
-// import `@docmost/prosemirror-markdown` natively — so this is the shared home.
 export function parseNodeArg(
   node: unknown,
   errMsg = "node was a string but not valid JSON",

packages/mcp/src/lib/transforms.ts

@@ -14,13 +14,13 @@
  *  - `marks` arrays are preserved verbatim when fragments are split/reordered.
  */
 
+import { blockPlainText } from "./node-ops.js";
+import { canonicalizeFootnotes } from "./footnote-canonicalize.js";
 import {
-  blockPlainText,
   footnoteContentKey,
   makeFootnoteDefinition,
   generateFootnoteId,
-} from "@docmost/prosemirror-markdown";
-import { canonicalizeFootnotes } from "./footnote-canonicalize.js";
+} from "./footnote-authoring.js";
 
 export { canonicalizeFootnotes } from "./footnote-canonicalize.js";
 
@@ -365,7 +365,7 @@ export function noteItem(inlineNodes: any[]): any {
  *   { type:"footnoteDefinition", attrs:{id}, content:[{ type:"paragraph", content }] }
  * (mirrors the editor-ext / docmost-schema FootnoteDefinition node).
  *
- * Built on the shared `makeFootnoteDefinition` factory (`@docmost/prosemirror-markdown`);
+ * Built on the shared `makeFootnoteDefinition` factory (footnote-authoring.ts);
  * the only extra is a fresh block id on the inner paragraph (Docmost stamps one,
  * and the canonicalizer preserves attrs as-is). Single factory, one place to
  * change the definition shape.

packages/mcp/src/tool-specs.ts

@@ -1115,4 +1115,112 @@ export const SHARED_TOOL_SPECS = {
       alt: z.string().optional(),
     }),
   },
+
+  // --- draw.io diagrams (issue #423, stage 1) ---
+
+  drawioGet: {
+    mcpName: 'drawio_get',
+    inAppKey: 'drawioGet',
+    description:
+      'Read a draw.io diagram on a page as mxGraph XML (default) or as its raw ' +
+      '`.drawio.svg`. `node` is the drawio node\'s attrs.id (from get_outline / ' +
+      'get_page_json) or "#<index>" for a top-level block. Returns the decoded ' +
+      'mxGraphModel XML plus meta { attachmentId, title, width, height, ' +
+      'cellCount, hash }. `hash` is the optimistic-lock key you MUST pass back ' +
+      'as baseHash to drawio_update. Diagrams a human saved from the editor ' +
+      '(including draw.io\'s compressed format) decode losslessly.',
+    tier: 'deferred',
+    catalogLine:
+      'drawioGet — read a draw.io diagram as mxGraph XML (+ hash for updates).',
+    buildShape: (z) => ({
+      pageId: z.string().min(1),
+      node: z
+        .string()
+        .min(1)
+        .describe('The drawio node attrs.id, or "#<index>" for a top-level block.'),
+      format: z
+        .enum(['xml', 'svg'])
+        .optional()
+        .describe('"xml" (default) for mxGraph XML, or "svg" for the raw .drawio.svg.'),
+    }),
+  },
+
+  drawioCreate: {
+    mcpName: 'drawio_create',
+    inAppKey: 'drawioCreate',
+    description:
+      'Create a draw.io diagram from mxGraph XML and insert it as a diagram ' +
+      'block. `xml` is a bare `<mxGraphModel>` OR a list of `<mxCell>` elements ' +
+      '(the server wraps it and adds the id=0 / id=1 sentinel cells). The XML is ' +
+      'LINTED first (well-formedness, sentinel cells, unique ids, vertex XOR ' +
+      'edge, every edge has a child <mxGeometry as="geometry"/>, edge ' +
+      'source/target and every parent resolve, style parses, no XML comments, ' +
+      'value escaping) — a violation returns a structured error naming the rule ' +
+      'and cellId so you can fix and retry. `where` positions the block like ' +
+      'insert_node: position before/after (with exactly one of anchorNodeId or ' +
+      'anchorText) or append. Returns { nodeId, attachmentId, warnings }. The ' +
+      'returned `nodeId` is an index-based "#<index>" handle (drawio nodes carry ' +
+      'no attrs.id): it addresses the new top-level block and can be fed straight ' +
+      'back into drawio_get / drawio_update for THIS document. It is positional, ' +
+      'so if you add or remove blocks before it, re-resolve via get_outline. The ' +
+      'diagram is editable in the draw.io editor and can be re-read with ' +
+      'drawio_get.',
+    tier: 'deferred',
+    catalogLine:
+      'drawioCreate — create a draw.io diagram from mxGraph XML and insert it.',
+    buildShape: (z) => ({
+      pageId: z.string().min(1),
+      xml: z
+        .string()
+        .min(1)
+        .describe(
+          'mxGraph XML: a bare <mxGraphModel> or a list of <mxCell> elements.',
+        ),
+      position: z
+        .enum(['before', 'after', 'append'])
+        .describe('Where to insert relative to the anchor.'),
+      anchorNodeId: z
+        .string()
+        .optional()
+        .describe('Anchor block id (for before/after).'),
+      anchorText: z
+        .string()
+        .optional()
+        .describe('Anchor text fragment (for before/after).'),
+      title: z.string().optional().describe('Optional diagram title.'),
+    }),
+  },
+
+  drawioUpdate: {
+    mcpName: 'drawio_update',
+    inAppKey: 'drawioUpdate',
+    description:
+      'Replace a draw.io diagram\'s content with new mxGraph XML (same lint ' +
+      'pipeline as drawio_create). `baseHash` is MANDATORY: pass the hash from ' +
+      'the drawio_get you based the edit on. If the diagram changed since ' +
+      '(a human or another agent edited it) the hash mismatches and the update ' +
+      'is refused with a conflict error — re-read with drawio_get and retry. On ' +
+      'success it overwrites the diagram attachment and updates the node ' +
+      'width/height. `node` is the drawio node attrs.id or "#<index>".',
+    tier: 'deferred',
+    catalogLine:
+      'drawioUpdate — replace a draw.io diagram (optimistic-locked by baseHash).',
+    buildShape: (z) => ({
+      pageId: z.string().min(1),
+      node: z
+        .string()
+        .min(1)
+        .describe('The drawio node attrs.id, or "#<index>" for a top-level block.'),
+      xml: z
+        .string()
+        .min(1)
+        .describe(
+          'New mxGraph XML: a bare <mxGraphModel> or a list of <mxCell> elements.',
+        ),
+      baseHash: z
+        .string()
+        .min(1)
+        .describe('The meta.hash from the drawio_get this edit is based on.'),
+    }),
+  },
 } satisfies Record<string, SharedToolSpec>;

packages/mcp/test/mock/drawio-tools.test.mjs

@@ -0,0 +1,467 @@
+// Contract tests for the drawio_get / drawio_create / drawio_update client
+// methods (issue #423). Follows the repo's seam-override pattern (see
+// full-doc-write-canonicalize.test.mjs): a DocmostClient subclass stubs the I/O
+// seams (auth, collab token, page read, attachment upload/fetch, the mutatePage
+// write) so the tool logic is exercised without a live Docmost or collab socket.
+import { test } from "node:test";
+import assert from "node:assert/strict";
+import pako from "pako";
+import { DocmostClient } from "../../build/client.js";
+import {
+  buildDrawioSvg,
+  encodeDrawioFile,
+  normalizeXml,
+  mxHash,
+  decodeDrawioSvg,
+} from "../../build/lib/drawio-xml.js";
+
+const MODEL =
+  '<mxGraphModel><root>' +
+  '<mxCell id="0"/><mxCell id="1" parent="0"/>' +
+  '<mxCell id="2" value="Hi" style="rounded=1;" vertex="1" parent="1">' +
+  '<mxGeometry x="20" y="20" width="120" height="60" as="geometry"/></mxCell>' +
+  '</root></mxGraphModel>';
+
+// Build a Docmost-style `.drawio.svg` (base64 content) for a model.
+function svgFor(model) {
+  return buildDrawioSvg(normalizeXml(model), "<g/>", { width: 200, height: 120 });
+}
+
+// Build a human/compressed-export `.drawio.svg` (base64 content wrapping a
+// compressed <diagram> payload), mimicking a diagram a person saved.
+function compressedSvgFor(model) {
+  const compressed = Buffer.from(
+    pako.deflateRaw(encodeURIComponent(normalizeXml(model))),
+  ).toString("base64");
+  const file = `<mxfile host="Electron"><diagram id="a" name="Page-1">${compressed}</diagram></mxfile>`;
+  const content = Buffer.from(file, "utf-8").toString("base64");
+  return `<svg xmlns="http://www.w3.org/2000/svg" content="${content}"><image href="x"/></svg>`;
+}
+
+// The vendored `drawio` node schema (diagramAttributes) declares ONLY these
+// attributes; PMNode.fromJSON drops anything else on save. Mirror that here so
+// the mock write path behaves like the real one — in particular, a block `id`
+// set on a drawio node does NOT survive the save, so a handle keyed on it is
+// un-resolvable. This is exactly what the production bug (issue #423 Fix 1) was.
+const DRAWIO_SCHEMA_ATTRS = new Set([
+  "src",
+  "title",
+  "alt",
+  "width",
+  "height",
+  "size",
+  "aspectRatio",
+  "align",
+  "attachmentId",
+]);
+
+function applyDrawioSchemaDrop(node) {
+  if (!node || typeof node !== "object") return;
+  if (node.type === "drawio" && node.attrs && typeof node.attrs === "object") {
+    for (const key of Object.keys(node.attrs)) {
+      if (!DRAWIO_SCHEMA_ATTRS.has(key)) delete node.attrs[key];
+    }
+  }
+  if (Array.isArray(node.content)) for (const c of node.content) applyDrawioSchemaDrop(c);
+}
+
+function makeClient({ pageDoc, attachmentSvg } = {}) {
+  const calls = { uploads: [], mutations: [] };
+  class TestClient extends DocmostClient {
+    async ensureAuthenticated() {}
+    async getCollabTokenWithReauth() {
+      return "collab-token";
+    }
+    async resolvePageId(pageId) {
+      return `uuid-${pageId}`;
+    }
+    async getPageRaw(pageId) {
+      return {
+        id: pageId,
+        slugId: "s",
+        title: "P",
+        spaceId: "sp",
+        content: pageDoc ?? { type: "doc", content: [] },
+      };
+    }
+    async uploadAttachmentBuffer(pageId, buffer, fileName, mime) {
+      const id = `att-${calls.uploads.length + 1}`;
+      calls.uploads.push({ pageId, fileName, mime, svg: buffer.toString("utf-8") });
+      return { id, fileName, fileSize: buffer.length };
+    }
+    async fetchAttachmentText(src) {
+      return attachmentSvg;
+    }
+    mutatePage(pageId, token, apiUrl, transform) {
+      // Run the transform against a clone of the source doc, capture the result.
+      const clone = structuredClone(pageDoc ?? { type: "doc", content: [] });
+      const doc = transform(clone);
+      // Mirror the real schema: unknown drawio attrs (e.g. a block `id`) are
+      // dropped on save, so callers can never rely on them to address the node.
+      if (doc) applyDrawioSchemaDrop(doc);
+      calls.mutations.push({ pageId, doc });
+      return Promise.resolve({ doc, verify: { changed: doc != null } });
+    }
+  }
+  const client = new TestClient("http://127.0.0.1:1/api", "e@x.com", "pw");
+  return { client, calls };
+}
+
+function findDrawio(node, acc = []) {
+  if (!node || typeof node !== "object") return acc;
+  if (node.type === "drawio") acc.push(node);
+  if (Array.isArray(node.content)) for (const c of node.content) findDrawio(c, acc);
+  return acc;
+}
+
+// --- drawio_create ---------------------------------------------------------
+
+test("drawio_create: lints, builds the .drawio.svg, uploads and inserts a node", async () => {
+  const pageDoc = {
+    type: "doc",
+    content: [{ type: "paragraph", attrs: { id: "p1" }, content: [] }],
+  };
+  const { client, calls } = makeClient({ pageDoc });
+  const res = await client.drawioCreate("page1", { position: "append" }, MODEL, "My diagram");
+
+  assert.equal(res.success, true);
+  // The returned handle is an index-based "#<index>" ref (drawio nodes carry no
+  // persisted attrs.id), addressing the appended top-level block (index 1, after
+  // the existing paragraph).
+  assert.equal(res.nodeId, "#1");
+  assert.equal(res.attachmentId, "att-1");
+  assert.equal(calls.uploads.length, 1);
+  assert.equal(calls.uploads[0].fileName, "diagram.drawio.svg");
+  assert.equal(calls.uploads[0].mime, "image/svg+xml");
+  // The uploaded SVG carries the model back (round-trips through the decode chain).
+  assert.equal(decodeDrawioSvg(calls.uploads[0].svg), normalizeXml(MODEL));
+
+  // A drawio node was appended with src/attachmentId/dimensions and the title.
+  const drawios = findDrawio(calls.mutations[0].doc);
+  assert.equal(drawios.length, 1);
+  const n = drawios[0];
+  // No `id` attribute is set/persisted on the node (schema has none).
+  assert.equal(n.attrs.id, undefined);
+  assert.equal(n.attrs.attachmentId, "att-1");
+  assert.match(n.attrs.src, /^\/api\/files\/att-1\//);
+  assert.ok(n.attrs.width > 0 && n.attrs.height > 0);
+  assert.equal(n.attrs.title, "My diagram");
+});
+
+test("drawio_create: a lint violation throws before any upload", async () => {
+  const { client, calls } = makeClient({ pageDoc: { type: "doc", content: [] } });
+  // Edge with no child geometry -> edge-geometry rule.
+  const bad =
+    '<mxGraphModel><root><mxCell id="0"/><mxCell id="1" parent="0"/>' +
+    '<mxCell id="2" vertex="1" parent="1"><mxGeometry x="0" y="0" width="10" height="10" as="geometry"/></mxCell>' +
+    '<mxCell id="3" edge="1" parent="1" source="2" target="2"/></root></mxGraphModel>';
+  await assert.rejects(
+    () => client.drawioCreate("page1", { position: "append" }, bad, undefined),
+    /edge-geometry/,
+  );
+  assert.equal(calls.uploads.length, 0, "no attachment uploaded on lint failure");
+});
+
+test("drawio_create: before/after requires exactly one anchor", async () => {
+  const { client } = makeClient({ pageDoc: { type: "doc", content: [] } });
+  await assert.rejects(
+    () => client.drawioCreate("page1", { position: "before" }, MODEL),
+    /exactly one of anchorNodeId or anchorText/,
+  );
+});
+
+// --- drawio_get ------------------------------------------------------------
+
+test("drawio_get: decodes the model and returns meta with a hash", async () => {
+  const pageDoc = {
+    type: "doc",
+    content: [
+      {
+        type: "drawio",
+        attrs: {
+          id: "d1",
+          src: "/api/files/att-1/diagram.drawio.svg",
+          attachmentId: "att-1",
+          title: "T",
+          width: 200,
+          height: 120,
+        },
+      },
+    ],
+  };
+  const { client } = makeClient({ pageDoc, attachmentSvg: svgFor(MODEL) });
+  const res = await client.drawioGet("page1", "d1", "xml");
+  assert.equal(res.content, normalizeXml(MODEL));
+  assert.equal(res.meta.attachmentId, "att-1");
+  assert.equal(res.meta.title, "T");
+  assert.equal(res.meta.cellCount, 1);
+  assert.equal(res.meta.hash, mxHash(normalizeXml(MODEL)));
+});
+
+test("drawio_get: format=svg returns the raw .drawio.svg", async () => {
+  const svg = svgFor(MODEL);
+  const pageDoc = {
+    type: "doc",
+    content: [
+      { type: "drawio", attrs: { id: "d1", src: "/api/files/att-1/x.svg", attachmentId: "att-1" } },
+    ],
+  };
+  const { client } = makeClient({ pageDoc, attachmentSvg: svg });
+  const res = await client.drawioGet("page1", "d1", "svg");
+  assert.equal(res.content, svg);
+});
+
+test("drawio_get: reads a HUMAN-saved compressed diagram losslessly (pako)", async () => {
+  const pageDoc = {
+    type: "doc",
+    content: [
+      { type: "drawio", attrs: { id: "d1", src: "/api/files/att-1/x.svg", attachmentId: "att-1" } },
+    ],
+  };
+  const { client } = makeClient({ pageDoc, attachmentSvg: compressedSvgFor(MODEL) });
+  const res = await client.drawioGet("page1", "d1", "xml");
+  assert.equal(res.content, normalizeXml(MODEL));
+});
+
+// --- drawio_update ---------------------------------------------------------
+
+const UPDATED_MODEL =
+  '<mxGraphModel><root>' +
+  '<mxCell id="0"/><mxCell id="1" parent="0"/>' +
+  '<mxCell id="2" value="Changed" style="rounded=1;" vertex="1" parent="1">' +
+  '<mxGeometry x="20" y="20" width="300" height="200" as="geometry"/></mxCell>' +
+  '</root></mxGraphModel>';
+
+function updatePageDoc() {
+  return {
+    type: "doc",
+    content: [
+      {
+        type: "drawio",
+        attrs: {
+          id: "d1",
+          src: "/api/files/att-1/diagram.drawio.svg",
+          attachmentId: "att-1",
+          width: 200,
+          height: 120,
+        },
+      },
+    ],
+  };
+}
+
+test("drawio_update: stale baseHash -> conflict, no upload", async () => {
+  const { client, calls } = makeClient({
+    pageDoc: updatePageDoc(),
+    attachmentSvg: svgFor(MODEL),
+  });
+  await assert.rejects(
+    () => client.drawioUpdate("page1", "d1", UPDATED_MODEL, "deadbeef-stale"),
+    /conflict/,
+  );
+  assert.equal(calls.uploads.length, 0, "no upload on conflict");
+});
+
+test("drawio_update: current baseHash -> uploads new attachment and repoints node dims", async () => {
+  const currentHash = mxHash(normalizeXml(MODEL));
+  const { client, calls } = makeClient({
+    pageDoc: updatePageDoc(),
+    attachmentSvg: svgFor(MODEL),
+  });
+  const res = await client.drawioUpdate("page1", "d1", UPDATED_MODEL, currentHash);
+  assert.equal(res.success, true);
+  assert.equal(res.attachmentId, "att-1"); // fresh id from the stub sequence
+  assert.equal(calls.uploads.length, 1);
+  // The uploaded SVG carries the NEW model.
+  assert.equal(decodeDrawioSvg(calls.uploads[0].svg), normalizeXml(UPDATED_MODEL));
+  // The node was repointed with the new bounding-box dimensions:
+  // vertex maxX=320,maxY=220 + the 20px preview margin -> 340 x 240.
+  const n = findDrawio(calls.mutations[0].doc)[0];
+  assert.equal(n.attrs.attachmentId, "att-1");
+  assert.equal(n.attrs.width, 340);
+  assert.equal(n.attrs.height, 240);
+  // The block `id` used as the legacy resolution handle is dropped on save
+  // (schema declares no `id`); the update still targeted the correct node.
+  assert.equal(n.attrs.id, undefined);
+});
+
+test("drawio_update: baseHash is mandatory", async () => {
+  const { client } = makeClient({ pageDoc: updatePageDoc(), attachmentSvg: svgFor(MODEL) });
+  await assert.rejects(
+    () => client.drawioUpdate("page1", "d1", UPDATED_MODEL, ""),
+    /baseHash is mandatory/,
+  );
+});
+
+// --- Fix 1: the create handle must resolve on the SAVED doc (no id) ---------
+
+test("drawio_create -> get/update: returned #<index> handle resolves on the saved doc (id dropped)", async () => {
+  // Create appends a drawio node after the existing paragraph.
+  const createDoc = {
+    type: "doc",
+    content: [{ type: "paragraph", attrs: { id: "p1" }, content: [] }],
+  };
+  const create = makeClient({ pageDoc: createDoc });
+  const res = await create.client.drawioCreate(
+    "page1",
+    { position: "append" },
+    MODEL,
+    "T",
+  );
+  // The handle is index-based, not a block id.
+  assert.equal(res.nodeId, "#1");
+
+  // Take the document EXACTLY as it was saved: the schema drop stripped the
+  // node's id, so no id-based handle could ever resolve against it.
+  const savedDoc = create.calls.mutations[0].doc;
+  assert.equal(findDrawio(savedDoc)[0].attrs.id, undefined);
+
+  // drawio_get with the returned handle resolves the just-created node.
+  const getClient = makeClient({ pageDoc: savedDoc, attachmentSvg: svgFor(MODEL) });
+  const got = await getClient.client.drawioGet("page1", res.nodeId, "xml");
+  assert.equal(got.nodeId, res.nodeId);
+  assert.equal(got.content, normalizeXml(MODEL));
+
+  // drawio_update with the same handle + the hash from get repoints that node.
+  const upClient = makeClient({ pageDoc: savedDoc, attachmentSvg: svgFor(MODEL) });
+  const upd = await upClient.client.drawioUpdate(
+    "page1",
+    res.nodeId,
+    UPDATED_MODEL,
+    got.meta.hash,
+  );
+  assert.equal(upd.success, true);
+  assert.equal(upd.nodeId, res.nodeId);
+  const updated = findDrawio(upClient.calls.mutations[0].doc)[0];
+  assert.equal(
+    decodeDrawioSvg(upClient.calls.uploads[0].svg),
+    normalizeXml(UPDATED_MODEL),
+  );
+  assert.equal(updated.attrs.width, 340);
+});
+
+// --- error paths: the LLM must get a clean error, not a crash --------------
+
+test("drawio_get: a bad node ref -> clean 'no node found' error", async () => {
+  // Page has one paragraph; the requested ref resolves to nothing.
+  const pageDoc = {
+    type: "doc",
+    content: [{ type: "paragraph", attrs: { id: "p1" }, content: [] }],
+  };
+  const { client } = makeClient({ pageDoc, attachmentSvg: svgFor(MODEL) });
+  await assert.rejects(
+    () => client.drawioGet("page1", "does-not-exist", "xml"),
+    /no node found for "does-not-exist"/,
+  );
+});
+
+test("drawio_get: a drawio node with no src -> clean 'has no src to read' error", async () => {
+  const pageDoc = {
+    type: "doc",
+    content: [
+      // A drawio node that carries no `src` (e.g. a half-written node).
+      { type: "drawio", attrs: { id: "d1", attachmentId: "att-1" } },
+    ],
+  };
+  const { client } = makeClient({ pageDoc, attachmentSvg: svgFor(MODEL) });
+  await assert.rejects(
+    () => client.drawioGet("page1", "d1", "xml"),
+    /node "d1" on page page1 has no src to read/,
+  );
+});
+
+test("drawio_update: the resolved node is NOT a drawio node -> clean error, no upload", async () => {
+  // "#0" resolves to a paragraph. The update must refuse cleanly rather than
+  // crash or repoint the wrong node.
+  const pageDoc = {
+    type: "doc",
+    content: [{ type: "paragraph", attrs: { id: "p1" }, content: [] }],
+  };
+  const { client, calls } = makeClient({ pageDoc, attachmentSvg: svgFor(MODEL) });
+  await assert.rejects(
+    () => client.drawioUpdate("page1", "#0", UPDATED_MODEL, "any-nonempty-hash"),
+    /node "#0" on page page1 is a paragraph, not a drawio diagram/,
+  );
+  assert.equal(calls.uploads.length, 0, "no upload when the node is not a diagram");
+  assert.equal(calls.mutations.length, 0, "no write when the node is not a diagram");
+});
+
+test("drawio_create: anchor not found -> clean error that reports the orphan attachment", async () => {
+  // The upload happens before the mutate transform; when the anchor cannot be
+  // found the write is skipped and the (now unreferenced) attachment is named
+  // in the error, exactly as the code documents.
+  const pageDoc = {
+    type: "doc",
+    content: [{ type: "paragraph", attrs: { id: "p1" }, content: [] }],
+  };
+  const { client, calls } = makeClient({ pageDoc });
+  await assert.rejects(
+    () =>
+      client.drawioCreate(
+        "page1",
+        { position: "after", anchorNodeId: "nope" },
+        MODEL,
+        "T",
+      ),
+    (err) =>
+      /anchor not found/.test(err.message) &&
+      /unreferenced orphan/.test(err.message) &&
+      /att-1/.test(err.message),
+  );
+  // The orphan was uploaded (and reported), but no node was written.
+  assert.equal(calls.uploads.length, 1, "attachment uploaded before the failed insert");
+  const drawios = calls.mutations.length ? findDrawio(calls.mutations[0].doc) : [];
+  assert.equal(drawios.length, 0, "no drawio node written when the anchor is missing");
+});
+
+// --- Fix 2: update targets ONLY the resolved node --------------------------
+
+test("drawio_update: repoints ONLY the addressed node, not siblings sharing an attachmentId", async () => {
+  // A copied diagram: two drawio nodes share one attachmentId. Updating via the
+  // "#0" handle must touch node #0 only, never the sibling copy.
+  const shared = {
+    type: "doc",
+    content: [
+      {
+        type: "drawio",
+        attrs: {
+          src: "/api/files/shared/x.svg",
+          attachmentId: "shared",
+          width: 200,
+          height: 120,
+        },
+      },
+      {
+        type: "drawio",
+        attrs: {
+          src: "/api/files/shared/x.svg",
+          attachmentId: "shared",
+          width: 200,
+          height: 120,
+        },
+      },
+    ],
+  };
+  const { client, calls } = makeClient({
+    pageDoc: shared,
+    attachmentSvg: svgFor(MODEL),
+  });
+  const res = await client.drawioUpdate(
+    "page1",
+    "#0",
+    UPDATED_MODEL,
+    mxHash(normalizeXml(MODEL)),
+  );
+  assert.equal(res.success, true);
+
+  const drawios = findDrawio(calls.mutations[0].doc);
+  assert.equal(drawios.length, 2);
+  // Node #0 repointed to the NEW attachment ("att-1" from the stub) and dims.
+  assert.equal(drawios[0].attrs.attachmentId, "att-1");
+  assert.equal(drawios[0].attrs.width, 340);
+  assert.match(drawios[0].attrs.src, /^\/api\/files\/att-1\//);
+  // Node #1 (the sibling copy) is untouched despite sharing the old attachmentId.
+  assert.equal(drawios[1].attrs.attachmentId, "shared");
+  assert.equal(drawios[1].attrs.width, 200);
+  assert.equal(drawios[1].attrs.src, "/api/files/shared/x.svg");
+});

packages/mcp/test/mock/footnote-warnings.test.mjs

@@ -1,7 +1,7 @@
 // Mock-HTTP test for the footnoteWarnings plumbing (#166). createPage is the
 // representative path that is fully plain-HTTP (import + getPage) and so is
 // mockable here; updatePage / importPageMarkdown attach footnoteWarnings with the
-// IDENTICAL wiring (`footnoteWarningsField(...)` spread-when-non-empty) but run their
+// IDENTICAL wiring (`analyzeFootnotes(...)` + spread-when-non-empty) but run their
 // mutation over the Hocuspocus collab WebSocket, which this plain-HTTP harness
 // does not stand up. The analyzer itself is unit-tested in footnote-analyze.test.
 import { test, after } from "node:test";
@@ -76,29 +76,35 @@ function pageHandler() {
   };
 }
 
-test("createPage attaches footnoteWarnings when the content uses legacy footnote syntax", async () => {
+test("createPage attaches footnoteWarnings when the content has footnote problems", async () => {
   const baseURL = await spawn(pageHandler());
   const client = new DocmostClient(baseURL, "user@example.com", "pw");
-  // Legacy reference-style `[^id]:` definitions — inert on import since #293.
-  const content = ["Intro[^a].", "", "[^a]: a definition"].join("\n");
+  // A dangling reference + a duplicate definition + a table marker.
+  const content = [
+    "Intro[^missing] and| cell[^t] |.",
+    "",
+    "[^d]: one",
+    "[^d]: two",
+    "[^t]: in table",
+  ].join("\n");
   const result = await client.createPage("T", content, "sp-1");
   assert.ok(Array.isArray(result.footnoteWarnings), "footnoteWarnings present");
   const joined = result.footnoteWarnings.join("\n");
-  assert.match(joined, /reference-style footnotes/i);
-  assert.match(joined, /\^\[footnote text\]/); // nudge to the inline form
+  assert.match(joined, /no matching definition/); // dangling [^missing]
+  assert.match(joined, /defined more than once/); // duplicate [^d]
   // The page itself is still returned.
   assert.equal(result.success, true);
 });
 
-test("createPage omits footnoteWarnings when the content uses the inline form", async () => {
+test("createPage omits footnoteWarnings when the content is clean", async () => {
   const baseURL = await spawn(pageHandler());
   const client = new DocmostClient(baseURL, "user@example.com", "pw");
-  const content = "A note.^[the body] and reuse.^[the body]";
+  const content = ["A[^a] and reuse[^a].", "", "[^a]: fine"].join("\n");
   const result = await client.createPage("T", content, "sp-1");
   assert.equal(
     "footnoteWarnings" in result,
     false,
-    "no footnoteWarnings field on inline-footnote input",
+    "no footnoteWarnings field on clean input",
   );
   assert.equal(result.success, true);
 });

packages/mcp/test/unit/drawio-preview.test.mjs

@@ -0,0 +1,84 @@
+// Unit tests for the pure-TS schematic SVG preview (issue #423). Asserts that
+// the preview emits well-formed SVG covering each primitive (rect/ellipse/
+// rhombus/edge), resolves container-relative coordinates to absolute, and that
+// the full `.drawio.svg` wrapper parses.
+import { test } from "node:test";
+import assert from "node:assert/strict";
+import { JSDOM } from "jsdom";
+import { renderDiagramShapes } from "../../build/lib/drawio-preview.js";
+import {
+  parseCells,
+  computeBBox,
+  buildDrawioSvg,
+  normalizeXml,
+} from "../../build/lib/drawio-xml.js";
+
+const { window } = new JSDOM("");
+function parseSvg(svg) {
+  const doc = new window.DOMParser().parseFromString(svg, "application/xml");
+  const err = doc.getElementsByTagName("parsererror");
+  assert.equal(err.length, 0, `SVG did not parse: ${err[0]?.textContent}`);
+  return doc;
+}
+
+const MODEL =
+  '<mxGraphModel><root>' +
+  '<mxCell id="0"/><mxCell id="1" parent="0"/>' +
+  '<mxCell id="2" value="Box &amp; Co" style="rounded=1;fillColor=#dae8fc;strokeColor=#6c8ebf;" vertex="1" parent="1">' +
+  '<mxGeometry x="40" y="40" width="120" height="60" as="geometry"/></mxCell>' +
+  '<mxCell id="3" value="Circle" style="ellipse;fillColor=#d5e8d4;" vertex="1" parent="1">' +
+  '<mxGeometry x="240" y="40" width="80" height="80" as="geometry"/></mxCell>' +
+  '<mxCell id="5" value="Dec" style="rhombus;" vertex="1" parent="1">' +
+  '<mxGeometry x="40" y="160" width="100" height="60" as="geometry"/></mxCell>' +
+  '<mxCell id="4" value="link" edge="1" parent="1" source="2" target="3"><mxGeometry relative="1" as="geometry"/></mxCell>' +
+  '</root></mxGraphModel>';
+
+test("renderDiagramShapes emits rect, ellipse, polygon and a line", () => {
+  const cells = parseCells(MODEL);
+  const inner = renderDiagramShapes(cells, computeBBox(cells));
+  assert.ok(inner.includes("<rect"), "has a rect");
+  assert.ok(inner.includes("<ellipse"), "has an ellipse");
+  assert.ok(inner.includes("<polygon"), "has a polygon (rhombus)");
+  assert.ok(inner.includes("<line"), "has an edge line");
+  // Labels are HTML-escaped.
+  assert.ok(inner.includes("Box &amp; Co"));
+});
+
+test("the full .drawio.svg wrapper parses as valid XML", () => {
+  const cells = parseCells(MODEL);
+  const bbox = computeBBox(cells);
+  const inner = renderDiagramShapes(cells, bbox);
+  const svg = buildDrawioSvg(normalizeXml(MODEL), inner, bbox);
+  const doc = parseSvg(svg);
+  assert.equal(doc.documentElement.tagName, "svg");
+  assert.ok(doc.documentElement.getAttribute("content"), "carries content=");
+  // The visible children exist.
+  assert.ok(doc.getElementsByTagName("rect").length >= 1);
+});
+
+test("container children resolve to absolute coordinates", () => {
+  // A group at (100,100) with a child rect at relative (10,10,20,20) -> abs 110,110.
+  const model =
+    '<mxGraphModel><root>' +
+    '<mxCell id="0"/><mxCell id="1" parent="0"/>' +
+    '<mxCell id="g" style="group;" vertex="1" parent="1"><mxGeometry x="100" y="100" width="200" height="200" as="geometry"/></mxCell>' +
+    '<mxCell id="c" value="in" vertex="1" parent="g"><mxGeometry x="10" y="10" width="20" height="20" as="geometry"/></mxCell>' +
+    '</root></mxGraphModel>';
+  const cells = parseCells(model);
+  const inner = renderDiagramShapes(cells, computeBBox(cells));
+  // The child rect must be placed at absolute x=110,y=110.
+  assert.ok(/<rect x="110" y="110"/.test(inner), `expected abs child rect, got: ${inner}`);
+});
+
+test("unknown stencil (shape=mxgraph.*) degrades to a labeled rectangle", () => {
+  const model =
+    '<mxGraphModel><root>' +
+    '<mxCell id="0"/><mxCell id="1" parent="0"/>' +
+    '<mxCell id="2" value="AWS" style="shape=mxgraph.aws4.lambda;" vertex="1" parent="1">' +
+    '<mxGeometry x="0" y="0" width="60" height="60" as="geometry"/></mxCell>' +
+    '</root></mxGraphModel>';
+  const cells = parseCells(model);
+  const inner = renderDiagramShapes(cells, computeBBox(cells));
+  assert.ok(inner.includes("<rect"), "unknown stencil -> rect");
+  assert.ok(inner.includes(">AWS<"), "keeps the label");
+});

packages/mcp/test/unit/drawio-xml.test.mjs

@@ -0,0 +1,360 @@
+// Unit tests for the drawio-xml module (issue #423): the linter (a positive
+// baseline + a negative case per rule, each asserting rule + cellId), the
+// decode chain (plain nested XML AND draw.io's compressed <diagram> via pako),
+// encode/round-trip byte-stability, hash stability, style parsing and the
+// bounding box.
+import { test } from "node:test";
+import assert from "node:assert/strict";
+import pako from "pako";
+import {
+  parseStyle,
+  lintModel,
+  prepareModel,
+  normalizeInput,
+  normalizeXml,
+  mxHash,
+  computeBBox,
+  parseCells,
+  decodeDrawioSvg,
+  decodeDrawioFileToModel,
+  buildDrawioSvg,
+  encodeDrawioFile,
+  countUserCells,
+  DrawioLintError,
+  inflateDiagramPayload,
+  MAX_INFLATED_DIAGRAM_BYTES,
+} from "../../build/lib/drawio-xml.js";
+
+// A well-formed model with one vertex and a valid edge to it.
+const VALID_MODEL =
+  '<mxGraphModel><root>' +
+  '<mxCell id="0"/><mxCell id="1" parent="0"/>' +
+  '<mxCell id="2" value="Hello" style="rounded=1;whiteSpace=wrap;html=1;" vertex="1" parent="1">' +
+  '<mxGeometry x="100" y="100" width="120" height="60" as="geometry"/></mxCell>' +
+  '<mxCell id="3" value="Two" style="ellipse;" vertex="1" parent="1">' +
+  '<mxGeometry x="300" y="100" width="80" height="80" as="geometry"/></mxCell>' +
+  '<mxCell id="4" edge="1" parent="1" source="2" target="3">' +
+  '<mxGeometry relative="1" as="geometry"/></mxCell>' +
+  '</root></mxGraphModel>';
+
+function issuesOf(fn) {
+  try {
+    fn();
+    return null;
+  } catch (e) {
+    assert.ok(e instanceof DrawioLintError, `expected DrawioLintError, got ${e}`);
+    return e.issues;
+  }
+}
+const hasRule = (issues, rule, cellId) =>
+  issues.some(
+    (i) => i.rule === rule && (cellId === undefined || i.cellId === cellId),
+  );
+
+// --- style parsing ---------------------------------------------------------
+
+test("parseStyle: base stylename + key=value pairs", () => {
+  const r = parseStyle("ellipse;fillColor=#ff0000;whiteSpace=wrap;");
+  assert.equal(r.baseStyle, "ellipse");
+  assert.equal(r.map.fillColor, "#ff0000");
+  assert.equal(r.map.whiteSpace, "wrap");
+  assert.equal(r.badSegment, undefined);
+});
+
+test("parseStyle: flags a segment with two '='", () => {
+  const r = parseStyle("a=b=c;");
+  assert.equal(r.badSegment, "a=b=c");
+});
+
+test("parseStyle: a second bare token is malformed", () => {
+  const r = parseStyle("rounded=1;bareword");
+  assert.equal(r.badSegment, "bareword");
+});
+
+// --- linter: positive baseline ---------------------------------------------
+
+test("lintModel: the canonical valid model passes", () => {
+  const { cells } = lintModel(VALID_MODEL);
+  assert.equal(cells.length, 5);
+});
+
+// --- linter: one negative case per rule ------------------------------------
+
+test("rule well-formed-xml: malformed XML", () => {
+  const issues = issuesOf(() => lintModel("<mxGraphModel><root><mxCell id=\"0\"></root>"));
+  assert.ok(hasRule(issues, "well-formed-xml"));
+  assert.ok(issues[0].position, "carries a line:col position");
+});
+
+test("rule structure: root is not mxGraphModel", () => {
+  const issues = issuesOf(() => lintModel("<foo><root/></foo>"));
+  assert.ok(hasRule(issues, "structure"));
+});
+
+test("rule sentinel-cells: missing id=0 / id=1", () => {
+  const m =
+    '<mxGraphModel><root>' +
+    '<mxCell id="2" vertex="1" parent="1"><mxGeometry x="0" y="0" width="10" height="10" as="geometry"/></mxCell>' +
+    '</root></mxGraphModel>';
+  const issues = issuesOf(() => lintModel(m));
+  assert.ok(hasRule(issues, "sentinel-cells", "0"));
+  assert.ok(hasRule(issues, "sentinel-cells", "1"));
+});
+
+test("rule duplicate-id: two cells share an id", () => {
+  const m =
+    '<mxGraphModel><root>' +
+    '<mxCell id="0"/><mxCell id="1" parent="0"/>' +
+    '<mxCell id="2" vertex="1" parent="1"><mxGeometry x="0" y="0" width="10" height="10" as="geometry"/></mxCell>' +
+    '<mxCell id="2" vertex="1" parent="1"><mxGeometry x="0" y="0" width="10" height="10" as="geometry"/></mxCell>' +
+    '</root></mxGraphModel>';
+  const issues = issuesOf(() => lintModel(m));
+  assert.ok(hasRule(issues, "duplicate-id", "2"));
+});
+
+test("rule vertex-edge-exclusive: cell is both vertex and edge", () => {
+  const m =
+    '<mxGraphModel><root>' +
+    '<mxCell id="0"/><mxCell id="1" parent="0"/>' +
+    '<mxCell id="2" vertex="1" edge="1" parent="1"><mxGeometry as="geometry"/></mxCell>' +
+    '</root></mxGraphModel>';
+  const issues = issuesOf(() => lintModel(m));
+  assert.ok(hasRule(issues, "vertex-edge-exclusive", "2"));
+});
+
+test("rule edge-geometry: self-closed edge without child mxGeometry", () => {
+  const m =
+    '<mxGraphModel><root>' +
+    '<mxCell id="0"/><mxCell id="1" parent="0"/>' +
+    '<mxCell id="2" vertex="1" parent="1"><mxGeometry x="0" y="0" width="10" height="10" as="geometry"/></mxCell>' +
+    '<mxCell id="3" vertex="1" parent="1"><mxGeometry x="30" y="0" width="10" height="10" as="geometry"/></mxCell>' +
+    '<mxCell id="4" edge="1" parent="1" source="2" target="3"/>' +
+    '</root></mxGraphModel>';
+  const issues = issuesOf(() => lintModel(m));
+  assert.ok(hasRule(issues, "edge-geometry", "4"));
+});
+
+test("rule edge-endpoint: source/target does not resolve", () => {
+  const m =
+    '<mxGraphModel><root>' +
+    '<mxCell id="0"/><mxCell id="1" parent="0"/>' +
+    '<mxCell id="2" vertex="1" parent="1"><mxGeometry x="0" y="0" width="10" height="10" as="geometry"/></mxCell>' +
+    '<mxCell id="4" edge="1" parent="1" source="2" target="99"><mxGeometry relative="1" as="geometry"/></mxCell>' +
+    '</root></mxGraphModel>';
+  const issues = issuesOf(() => lintModel(m));
+  assert.ok(hasRule(issues, "edge-endpoint", "4"));
+});
+
+test("rule parent-exists: parent points at a missing id", () => {
+  const m =
+    '<mxGraphModel><root>' +
+    '<mxCell id="0"/><mxCell id="1" parent="0"/>' +
+    '<mxCell id="2" vertex="1" parent="42"><mxGeometry x="0" y="0" width="10" height="10" as="geometry"/></mxCell>' +
+    '</root></mxGraphModel>';
+  const issues = issuesOf(() => lintModel(m));
+  assert.ok(hasRule(issues, "parent-exists", "2"));
+});
+
+test("rule no-comments: XML comment present", () => {
+  const m =
+    '<mxGraphModel><root>' +
+    '<!-- a comment --><mxCell id="0"/><mxCell id="1" parent="0"/>' +
+    '</root></mxGraphModel>';
+  const issues = issuesOf(() => lintModel(m));
+  assert.ok(hasRule(issues, "no-comments"));
+});
+
+test("rule style-format: malformed style segment (cellId reported)", () => {
+  const m =
+    '<mxGraphModel><root>' +
+    '<mxCell id="0"/><mxCell id="1" parent="0"/>' +
+    '<mxCell id="2" style="rounded=1;a=b=c;" vertex="1" parent="1"><mxGeometry x="0" y="0" width="10" height="10" as="geometry"/></mxCell>' +
+    '</root></mxGraphModel>';
+  const issues = issuesOf(() => lintModel(m));
+  assert.ok(hasRule(issues, "style-format", "2"));
+});
+
+test("rule value-newline: literal newline in a value (cellId reported)", () => {
+  const m =
+    '<mxGraphModel><root>' +
+    '<mxCell id="0"/><mxCell id="1" parent="0"/>' +
+    '<mxCell id="2" value="line1\nline2" vertex="1" parent="1"><mxGeometry x="0" y="0" width="10" height="10" as="geometry"/></mxCell>' +
+    '</root></mxGraphModel>';
+  const issues = issuesOf(() => lintModel(m));
+  assert.ok(hasRule(issues, "value-newline", "2"));
+});
+
+test("rule value-escaping: unescaped ampersand in a value (cellId reported)", () => {
+  const m =
+    '<mxGraphModel><root>' +
+    '<mxCell id="0"/><mxCell id="1" parent="0"/>' +
+    '<mxCell id="2" value="A & B" vertex="1" parent="1"><mxGeometry x="0" y="0" width="10" height="10" as="geometry"/></mxCell>' +
+    '</root></mxGraphModel>';
+  const issues = issuesOf(() => lintModel(m));
+  assert.ok(hasRule(issues, "value-escaping", "2"));
+});
+
+test("rule reserved id: escaped entity value passes (no false positive)", () => {
+  const m =
+    '<mxGraphModel><root>' +
+    '<mxCell id="0"/><mxCell id="1" parent="0"/>' +
+    '<mxCell id="2" value="A &amp; B &lt;ok&gt;" vertex="1" parent="1"><mxGeometry x="0" y="0" width="10" height="10" as="geometry"/></mxCell>' +
+    '</root></mxGraphModel>';
+  assert.doesNotThrow(() => lintModel(m));
+});
+
+// --- input normalization ---------------------------------------------------
+
+test("normalizeInput: a list of <mxCell> is wrapped and sentinels added", () => {
+  const frag =
+    '<mxCell id="2" vertex="1" parent="1"><mxGeometry x="0" y="0" width="10" height="10" as="geometry"/></mxCell>';
+  const model = normalizeInput(frag);
+  assert.ok(model.startsWith("<mxGraphModel"));
+  assert.ok(model.includes('<mxCell id="0"/>'));
+  assert.ok(model.includes('<mxCell id="1" parent="0"/>'));
+  // And it lints clean.
+  assert.doesNotThrow(() => lintModel(model));
+});
+
+test("normalizeInput: an existing sentinel is not duplicated", () => {
+  const frag =
+    '<mxCell id="0"/><mxCell id="1" parent="0"/>' +
+    '<mxCell id="2" vertex="1" parent="1"><mxGeometry x="0" y="0" width="10" height="10" as="geometry"/></mxCell>';
+  const model = normalizeInput(frag);
+  const count0 = (model.match(/id="0"/g) || []).length;
+  assert.equal(count0, 1);
+});
+
+test("prepareModel: returns bbox, cellCount, hash and lints", () => {
+  const p = prepareModel(VALID_MODEL);
+  assert.equal(p.cellCount, 3); // 2, 3, 4 (sentinels excluded)
+  assert.ok(p.bbox.width > 0 && p.bbox.height > 0);
+  assert.equal(p.hash, mxHash(normalizeXml(VALID_MODEL)));
+});
+
+// --- decode chain: plain -----------------------------------------------------
+
+test("decode chain (plain): buildDrawioSvg -> decodeDrawioSvg round-trips byte-stable", () => {
+  const model = normalizeXml(VALID_MODEL);
+  const svg = buildDrawioSvg(model, "<g/>", { width: 400, height: 200 });
+  const decoded = decodeDrawioSvg(svg);
+  assert.equal(decoded, model);
+});
+
+test("decode chain: entity-encoded content= (draw.io export style) is read directly", () => {
+  const model = normalizeXml(VALID_MODEL);
+  const file = encodeDrawioFile(model, "Page-1");
+  const escaped = file
+    .replace(/&/g, "&amp;")
+    .replace(/</g, "&lt;")
+    .replace(/>/g, "&gt;")
+    .replace(/"/g, "&quot;");
+  const svg = `<svg xmlns="http://www.w3.org/2000/svg" content="${escaped}"></svg>`;
+  const decoded = decodeDrawioSvg(svg);
+  assert.equal(decoded, model);
+});
+
+// --- decode chain: compressed (pako) ---------------------------------------
+
+test("decode chain (compressed pako): human-saved <diagram> payload decodes losslessly", () => {
+  const model = normalizeXml(VALID_MODEL);
+  // Reproduce draw.io's compression: encodeURIComponent -> raw deflate -> base64.
+  const compressed = Buffer.from(
+    pako.deflateRaw(encodeURIComponent(model)),
+  ).toString("base64");
+  const file = `<mxfile host="Electron"><diagram id="abc" name="Page-1">${compressed}</diagram></mxfile>`;
+  // Docmost stores the file base64 in content=.
+  const contentB64 = Buffer.from(file, "utf-8").toString("base64");
+  const svg = `<svg xmlns="http://www.w3.org/2000/svg" content="${contentB64}"><image href="x"/></svg>`;
+  const decoded = decodeDrawioSvg(svg);
+  assert.equal(decoded, model);
+});
+
+test("decodeDrawioFileToModel: bare mxGraphModel file returns the model substring", () => {
+  const model = normalizeXml(VALID_MODEL);
+  assert.equal(decodeDrawioFileToModel(model), model);
+});
+
+// --- hash stability --------------------------------------------------------
+
+test("mxHash: stable across inter-tag whitespace, sensitive to content", () => {
+  const a = VALID_MODEL;
+  const b = VALID_MODEL.replace(/></g, ">\n  <"); // reformat only
+  assert.equal(mxHash(a), mxHash(b));
+  const c = VALID_MODEL.replace('value="Hello"', 'value="Changed"');
+  assert.notEqual(mxHash(a), mxHash(c));
+});
+
+// --- bounding box + cell count ---------------------------------------------
+
+test("computeBBox + countUserCells", () => {
+  const cells = parseCells(VALID_MODEL);
+  const bbox = computeBBox(cells);
+  // Vertex 3 spans to x=380,y=180; plus the 20px margin.
+  assert.equal(bbox.width, 400);
+  assert.equal(bbox.height, 200);
+  assert.equal(countUserCells(VALID_MODEL), 3);
+});
+
+// --- decompression-bomb guard (Fix 3) --------------------------------------
+
+test("inflateDiagramPayload: a small legitimate payload inflates fine", () => {
+  const xml = normalizeXml(VALID_MODEL);
+  const base64 = Buffer.from(
+    pako.deflateRaw(encodeURIComponent(xml)),
+  ).toString("base64");
+  assert.equal(inflateDiagramPayload(base64), xml);
+});
+
+test("inflateDiagramPayload: rejects an over-cap decompression bomb", () => {
+  // A tiny compressed payload that inflates to just over the cap. Highly
+  // compressible (all one byte) -> the base64 is small, but the inflated output
+  // exceeds MAX_INFLATED_DIAGRAM_BYTES and must be refused before it is fully
+  // materialised.
+  const bombSize = MAX_INFLATED_DIAGRAM_BYTES + 1024;
+  const base64 = Buffer.from(
+    pako.deflateRaw(Buffer.alloc(bombSize, 0x41 /* 'A' */)),
+  ).toString("base64");
+  assert.ok(
+    base64.length < 1024 * 1024,
+    "the compressed bomb is tiny relative to its inflated size",
+  );
+  assert.throws(
+    () => inflateDiagramPayload(base64),
+    /decompression bomb/,
+  );
+});
+
+test("encode/build: a title with < > \" & round-trips without corrupting the SVG", async () => {
+  // A user-supplied title full of XML metacharacters must be escaped so the
+  // inner <mxfile> stays well-formed and the outer content="..." attribute is
+  // never broken out of. Prove it survives the encode -> build -> decode chain.
+  const title = 'A < B > C " D & E';
+  const model = normalizeXml(VALID_MODEL);
+  const svg = buildDrawioSvg(model, "<g/>", { width: 200, height: 120 }, title);
+
+  // The outer content="..." attribute must not be broken by the title: the raw
+  // title metacharacters never appear literally in the SVG markup (they are
+  // base64-encoded inside content=, and escaped inside the file XML).
+  const contentMatch = /content="([^"]*)"/.exec(svg);
+  assert.ok(contentMatch, "SVG has a single well-formed content= attribute");
+
+  // The diagram model still decodes losslessly despite the exotic title.
+  assert.equal(decodeDrawioSvg(svg), model);
+
+  // The file XML is well-formed: the title lives in name="..." as escaped
+  // entities, so unescaping recovers the original title byte-for-byte.
+  const fileXml = Buffer.from(contentMatch[1], "base64").toString("utf-8");
+  const nameMatch = /<diagram id="[^"]*" name="([^"]*)">/.exec(fileXml);
+  assert.ok(nameMatch, "the diagram name attribute is intact and quote-safe");
+  const decodedTitle = nameMatch[1]
+    .replace(/&lt;/g, "<")
+    .replace(/&gt;/g, ">")
+    .replace(/&quot;/g, '"')
+    .replace(/&amp;/g, "&");
+  assert.equal(decodedTitle, title);
+
+  // encodeDrawioFile alone produces the same escaped, well-formed envelope.
+  const file = encodeDrawioFile(model, title);
+  assert.match(file, /name="A &lt; B &gt; C &quot; D &amp; E">/);
+});

packages/mcp/test/unit/footnote-analyze.test.mjs

@@ -1,45 +1,64 @@
 import { test } from "node:test";
 import assert from "node:assert/strict";
 
-import {
-  footnoteWarningsField,
-  hasLegacyFootnoteDefinition,
-} from "../../build/lib/footnote-analyze.js";
+import { analyzeFootnotes } from "../../build/lib/footnote-analyze.js";
 
-// #414: the legacy footnote diagnostics were reduced to ONE advisory that fires
-// on the PRESENCE of legacy reference-style `[^id]:` definition syntax (inert on
-// import since #293), nudging the author to inline `^[...]` footnotes.
-
-test("inline `^[...]` footnotes produce no warning", () => {
-  const md = "A note here.^[the body] and reuse elsewhere.^[the body]";
-  assert.equal(hasLegacyFootnoteDefinition(md), false);
-  assert.deepEqual(footnoteWarningsField(md), {});
+test("clean footnotes produce no diagnostics", () => {
+  const md = ["A[^a] and B[^b].", "", "[^a]: first", "[^b]: second"].join("\n");
+  const d = analyzeFootnotes(md);
+  assert.deepEqual(d.danglingReferences, []);
+  assert.deepEqual(d.emptyDefinitions, []);
+  assert.deepEqual(d.duplicateDefinitions, []);
+  assert.deepEqual(d.referencesInTables, []);
+  assert.deepEqual(d.warnings, []);
 });
 
-test("no footnotes at all produce no warning", () => {
-  const md = "Just a paragraph with [a link](https://x) and no footnotes.";
-  assert.equal(hasLegacyFootnoteDefinition(md), false);
-  assert.deepEqual(footnoteWarningsField(md), {});
+test("reuse (repeated references to one definition) is NOT a warning", () => {
+  const md = ["A[^a] B[^a] C[^a].", "", "[^a]: shared"].join("\n");
+  const d = analyzeFootnotes(md);
+  assert.deepEqual(d.danglingReferences, []);
+  assert.deepEqual(d.warnings, []);
 });
 
-test("a legacy `[^id]:` definition triggers the single advisory", () => {
-  const md = ["See[^a].", "", "[^a]: defined"].join("\n");
-  assert.equal(hasLegacyFootnoteDefinition(md), true);
-  const field = footnoteWarningsField(md);
-  assert.equal(field.footnoteWarnings.length, 1);
-  assert.match(field.footnoteWarnings[0], /reference-style footnotes/i);
-  assert.match(field.footnoteWarnings[0], /\^\[footnote text\]/);
+test("dangling reference (no definition) is reported", () => {
+  const md = ["See[^missing] and[^a].", "", "[^a]: defined"].join("\n");
+  const d = analyzeFootnotes(md);
+  assert.deepEqual(d.danglingReferences, ["missing"]);
+  assert.equal(d.warnings.length, 1);
+  assert.match(d.warnings[0], /no matching definition/);
+  assert.match(d.warnings[0], /\[\^missing\]/);
 });
 
-test("a bare `[^id]` reference (no definition line) is not flagged", () => {
-  // Only the definition syntax `[^id]:` is a reliable signal of legacy authoring;
-  // a lone `[^x]` in prose is too ambiguous to warn on.
-  const md = "A sentence mentioning [^x] with no definition.";
-  assert.equal(hasLegacyFootnoteDefinition(md), false);
-  assert.deepEqual(footnoteWarningsField(md), {});
+test("empty definition text is reported", () => {
+  const md = ["See[^a].", "", "[^a]:   "].join("\n");
+  const d = analyzeFootnotes(md);
+  assert.deepEqual(d.emptyDefinitions, ["a"]);
+  assert.match(d.warnings.join("\n"), /empty text/);
 });
 
-test("legacy syntax inside a code fence is ignored (fence-aware)", () => {
+test("duplicate definition id is reported (first-wins)", () => {
+  const md = ["See[^d].", "", "[^d]: first", "[^d]: second"].join("\n");
+  const d = analyzeFootnotes(md);
+  assert.deepEqual(d.duplicateDefinitions, ["d"]);
+  assert.match(d.warnings.join("\n"), /defined more than once/);
+});
+
+test("reference inside a GFM table row is reported (heuristic)", () => {
+  const md = [
+    "| Col |",
+    "| --- |",
+    "| cell[^t] |",
+    "",
+    "[^t]: table note",
+  ].join("\n");
+  const d = analyzeFootnotes(md);
+  assert.deepEqual(d.referencesInTables, ["t"]);
+  assert.match(d.warnings.join("\n"), /table/);
+  // It is defined, so it is NOT also dangling.
+  assert.deepEqual(d.danglingReferences, []);
+});
+
+test("footnote syntax inside a code fence is ignored", () => {
   const md = [
     "Intro.",
     "",
@@ -48,22 +67,40 @@ test("legacy syntax inside a code fence is ignored (fence-aware)", () => {
     "[^demo]: not a real definition",
     "```",
     "",
-    "Outro with an inline note.^[real]",
+    "Outro[^a].",
+    "",
+    "[^a]: real",
   ].join("\n");
-  assert.equal(hasLegacyFootnoteDefinition(md), false);
-  assert.deepEqual(footnoteWarningsField(md), {});
+  const d = analyzeFootnotes(md);
+  // `[^demo]` lives only in the fenced block, so it is neither a reference nor a
+  // dangling one, and `[^demo]:` is not counted as a definition.
+  assert.deepEqual(d.danglingReferences, []);
+  assert.deepEqual(d.duplicateDefinitions, []);
+  assert.deepEqual(d.warnings, []);
 });
 
-test("a legacy definition OUTSIDE a fence still warns even with a fenced sample", () => {
-  const md = [
-    "```",
-    "[^demo]: example inside a fence",
-    "```",
-    "",
-    "See[^a].",
-    "",
-    "[^a]: real definition outside the fence",
-  ].join("\n");
-  assert.equal(hasLegacyFootnoteDefinition(md), true);
-  assert.equal(footnoteWarningsField(md).footnoteWarnings.length, 1);
+test("a reference that only appears inside a definition's text is not dangling", () => {
+  // `[^b]` is referenced from within [^a]'s text and has its own definition.
+  const md = ["See[^a].", "", "[^a]: see also [^b]", "[^b]: the other"].join(
+    "\n",
+  );
+  const d = analyzeFootnotes(md);
+  assert.deepEqual(d.danglingReferences, []);
+});
+
+test("multiple problem classes accumulate distinct warnings", () => {
+  const md = [
+    "Ref[^x] and[^dup].",
+    "",
+    "[^dup]: one",
+    "[^dup]: two",
+    "[^empty]:",
+  ].join("\n");
+  const d = analyzeFootnotes(md);
+  // x has no definition; dup is defined twice; empty is empty AND has no ref.
+  assert.ok(d.danglingReferences.includes("x"));
+  assert.deepEqual(d.duplicateDefinitions, ["dup"]);
+  assert.deepEqual(d.emptyDefinitions, ["empty"]);
+  // One warning line per problem class present.
+  assert.ok(d.warnings.length >= 3);
 });

packages/mcp/test/unit/footnote-canonicalize.test.mjs

@@ -5,7 +5,7 @@ import { canonicalizeFootnotes } from "../../build/lib/footnote-canonicalize.js"
 import {
   footnoteContentKey,
   generateFootnoteId,
-} from "@docmost/prosemirror-markdown";
+} from "../../build/lib/footnote-authoring.js";
 import { insertInlineFootnote } from "../../build/lib/transforms.js";
 import { markdownToProseMirrorCanonical } from "../../build/lib/collaboration.js";
 

packages/mcp/test/unit/footnote-warnings-import.test.mjs

@@ -1,37 +1,39 @@
 import { test } from "node:test";
 import assert from "node:assert/strict";
 
-import { footnoteWarningsField } from "../../build/lib/footnote-analyze.js";
+import {
+  analyzeFootnotes,
+  footnoteWarningsField,
+} from "../../build/lib/footnote-analyze.js";
 import {
   serializeDocmostMarkdown,
   parseDocmostMarkdown,
 } from "../../build/lib/markdown-document.js";
 
-// Pins the footnoteWarnings PLUMBING contract (#169 review; reduced in #414): the
-// field is present only when legacy reference-style `[^id]:` syntax is used and
-// omitted otherwise, AND `import_page_markdown` analyzes the BODY (after the
-// docmost:meta / docmost:comments blocks) — so a footnote-like token inside those
-// JSON blocks never warns, while a real definition in the body does.
-// importPageMarkdown does exactly `footnoteWarningsField(parseDocmostMarkdown(full).body)`
-// over a collab socket this harness does not stand up, so we test the same pure
-// composition directly.
+// Pins the footnoteWarnings PLUMBING contract (#169 review): the field is
+// present only on problems and omitted on clean input, AND `import_page_markdown`
+// analyzes the BODY (after the docmost:meta / docmost:comments blocks) — so a
+// footnote-like token inside those JSON blocks never warns, while a real marker
+// in the body does. importPageMarkdown does exactly
+// `footnoteWarningsField(parseDocmostMarkdown(full).body)` over a collab socket
+// this harness does not stand up, so we test the same pure composition directly.
 
-test("footnoteWarningsField is present on legacy syntax and omitted on the inline form", () => {
-  const legacy = footnoteWarningsField("See[^a].\n\n[^a]: defined");
-  assert.ok(Array.isArray(legacy.footnoteWarnings));
-  assert.match(legacy.footnoteWarnings.join("\n"), /reference-style footnotes/i);
+test("footnoteWarningsField is present on problems and omitted on clean input", () => {
+  const problem = footnoteWarningsField("See[^missing].\n\n[^a]: defined");
+  assert.ok(Array.isArray(problem.footnoteWarnings));
+  assert.match(problem.footnoteWarnings.join("\n"), /no matching definition/);
 
-  const inline = footnoteWarningsField("A note.^[the body] reused.^[the body]");
-  assert.deepEqual(inline, {}); // no key at all on inline-footnote input
+  const clean = footnoteWarningsField("A[^a] and reuse[^a].\n\n[^a]: fine");
+  assert.deepEqual(clean, {}); // no key at all on clean input
 });
 
 test("import analyzes the BODY only — tokens inside meta/comments never warn", () => {
-  // meta + comments JSON carry `[^metaonly]:` / `[^commentonly]:`-looking text;
-  // the BODY has a genuine legacy `[^bodyref]:` definition.
+  // meta + comments JSON carry `[^metaonly]` / `[^commentonly]`-looking text; the
+  // BODY has a genuinely dangling `[^bodyref]`.
   const full = serializeDocmostMarkdown(
-    { pageId: "p1", note: "front-matter mentions [^metaonly]: in text" },
-    "Body with a legacy[^bodyref] marker.\n\n[^bodyref]: the definition",
-    [{ id: "c1", content: "a comment that says [^commentonly]: text" }],
+    { pageId: "p1", note: "front-matter mentions [^metaonly] in text" },
+    "Body with a dangling[^bodyref] marker.",
+    [{ id: "c1", content: "a comment that says [^commentonly]" }],
   );
 
   const { body } = parseDocmostMarkdown(full);
@@ -40,19 +42,20 @@ test("import analyzes the BODY only — tokens inside meta/comments never warn",
   assert.ok(!body.includes("[^commentonly]"));
 
   const field = footnoteWarningsField(body);
-  // ONLY the body's legacy definition triggers the advisory.
-  assert.ok(Array.isArray(field.footnoteWarnings));
-  assert.match(field.footnoteWarnings.join("\n"), /reference-style footnotes/i);
+  const joined = (field.footnoteWarnings ?? []).join("\n");
+  // ONLY the body's dangling reference is flagged.
+  assert.match(joined, /\[\^bodyref\]/);
+  assert.ok(!joined.includes("metaonly"));
+  assert.ok(!joined.includes("commentonly"));
 
-  // The meta/comments tokens, analyzed on their own, would NOT have warned in a
-  // way that leaks here — the field is computed over the body only.
-  assert.deepEqual(footnoteWarningsField("front-matter mentions text"), {});
+  // Cross-check against analyzeFootnotes directly (same composition the importer uses).
+  assert.deepEqual(analyzeFootnotes(body).danglingReferences, ["bodyref"]);
 });
 
-test("import on an inline-footnote body yields no footnoteWarnings field", () => {
+test("import on a clean body yields no footnoteWarnings field", () => {
   const full = serializeDocmostMarkdown(
     { pageId: "p1" },
-    "Clean body.^[a note] reusing.^[a note]",
+    "Clean body[^a] reusing[^a].\n\n[^a]: ok",
     [],
   );
   const { body } = parseDocmostMarkdown(full);

packages/mcp/test/unit/node-ops-table.test.mjs

@@ -5,7 +5,7 @@ import {
   insertNodeRelative,
   sanitizeForYjs,
   findUnstorableAttr,
-} from "@docmost/prosemirror-markdown";
+} from "../../build/lib/node-ops.js";
 
 // ProseMirror builders. Blocks carry a stable id in attrs.id.
 const textNode = (text) => ({ type: "text", text });

packages/mcp/test/unit/node-ops.test.mjs

@@ -7,7 +7,7 @@ import {
   deleteNodeById,
   assertUnambiguousMatch,
   insertNodeRelative,
-} from "@docmost/prosemirror-markdown";
+} from "../../build/lib/node-ops.js";
 
 // ProseMirror builders. Blocks carry a stable id in attrs.id.
 const textNode = (text) => ({ type: "text", text });

packages/mcp/test/unit/outline.test.mjs

@@ -1,7 +1,7 @@
 import { test } from "node:test";
 import assert from "node:assert/strict";
 
-import { buildOutline, getNodeByRef } from "@docmost/prosemirror-markdown";
+import { buildOutline, getNodeByRef } from "../../build/lib/node-ops.js";
 
 // Helpers to build the small fixture doc.
 const textNode = (text) => ({ type: "text", text });

packages/mcp/test/unit/page-search.test.mjs

@@ -2,7 +2,7 @@ import { test } from "node:test";
 import assert from "node:assert/strict";
 
 import { searchInDoc } from "../../build/lib/page-search.js";
-import { getNodeByRef } from "@docmost/prosemirror-markdown";
+import { getNodeByRef } from "../../build/lib/node-ops.js";
 
 // ---------------------------------------------------------------------------
 // Document builders. Mirror the Docmost ProseMirror shape: paragraphs/headings

packages/mcp/test/unit/parse-node-arg.test.mjs

@@ -1,7 +1,7 @@
 import { test } from "node:test";
 import assert from "node:assert/strict";
 
-import { parseNodeArg } from "@docmost/prosemirror-markdown";
+import { parseNodeArg } from "../../build/lib/parse-node-arg.js";
 
 test("parseNodeArg passes an object through unchanged", () => {
   const obj = { type: "paragraph", content: [] };

packages/mcp/test/unit/table-ops.test.mjs

@@ -6,7 +6,7 @@ import {
   insertTableRow,
   deleteTableRow,
   updateTableCell,
-} from "@docmost/prosemirror-markdown";
+} from "../../build/lib/node-ops.js";
 
 // ---------------------------------------------------------------------------
 // Builders. Tables/rows/cells carry NO attrs.id — only the paragraph inside a

packages/prosemirror-markdown/src/lib/footnote.ts

@@ -59,89 +59,3 @@ export function splitFootnoteParagraphs(encoded: string): string[] {
   paragraphs.push(current);
   return paragraphs;
 }
-
-// ---------------------------------------------------------------------------
-// Inline-authoring helpers (#414: moved here from the mcp `footnote-authoring.ts`
-// fork so the dedup convention — content-key + definition factory + id gen —
-// has ONE home next to the importer that shares the convention). Used by the
-// mcp author-inline tool (`insertInlineFootnote` in transforms.ts).
-// ---------------------------------------------------------------------------
-
-const FOOTNOTE_DEFINITION_NAME = "footnoteDefinition";
-
-function cloneJson<T>(v: T): T {
-  if (typeof structuredClone === "function") return structuredClone(v);
-  return JSON.parse(JSON.stringify(v)) as T;
-}
-
-/**
- * Normalized content key for de-duplicating footnote DEFINITIONS by their text.
- *
- * Two definitions with the same key are the SAME footnote — so the inline
- * authoring tool reuses one id (one number, one definition, several references)
- * instead of minting a second definition. Key = plaintext (whitespace-collapsed,
- * trimmed) PLUS a signature of the inline mark types in order, so two notes that
- * read the same but differ in formatting (one bold, one plain) are NOT merged.
- * Conservative: only an exact match merges.
- */
-export function footnoteContentKey(defNode: any): string {
-  const parts: string[] = [];
-  const visit = (n: any): void => {
-    if (!n || typeof n !== "object") return;
-    if (n.type === "text" && typeof n.text === "string") {
-      const marks = Array.isArray(n.marks)
-        ? n.marks.map((m: any) => m?.type).filter(Boolean).sort().join(",")
-        : "";
-      parts.push(`${n.text}${marks}`);
-    }
-    if (Array.isArray(n.content)) for (const c of n.content) visit(c);
-  };
-  visit(defNode);
-  // Collapse the assembled text's whitespace and trim, keeping the mark
-  // signature attached so formatting differences still distinguish notes.
-  return parts
-    .join("")
-    .replace(/[ \t\r\n]+/g, " ")
-    .trim();
-}
-
-/**
- * Build a footnoteDefinition node from inline ProseMirror nodes, keyed by id.
- */
-export function makeFootnoteDefinition(id: string, inlineNodes: any[]): any {
-  const content = Array.isArray(inlineNodes) ? cloneJson(inlineNodes) : [];
-  return {
-    type: FOOTNOTE_DEFINITION_NAME,
-    attrs: { id },
-    content: [{ type: "paragraph", content }],
-  };
-}
-
-/**
- * Generate a uuidv7-style id (time-ordered), matching editor-ext's
- * `generateFootnoteId`. Used for a genuinely-new inline footnote id.
- */
-export function generateFootnoteId(): string {
-  const now = Date.now();
-  const timeHex = now.toString(16).padStart(12, "0");
-  const rand = (length: number) => {
-    let s = "";
-    for (let i = 0; i < length; i++)
-      s += Math.floor(Math.random() * 16).toString(16);
-    return s;
-  };
-  const versioned = "7" + rand(3);
-  const variantNibble = (8 + Math.floor(Math.random() * 4)).toString(16);
-  const variant = variantNibble + rand(3);
-  return (
-    timeHex.slice(0, 8) +
-    "-" +
-    timeHex.slice(8, 12) +
-    "-" +
-    versioned +
-    "-" +
-    variant +
-    "-" +
-    rand(12)
-  );
-}

packages/prosemirror-markdown/src/lib/index.ts

@@ -44,35 +44,3 @@ export {
   docsCanonicallyEqual,
 } from "./canonicalize.js";
 export { parsePageFile, serializePageFile } from "./page-file.js";
-
-// Pure, network-free helpers for manipulating a ProseMirror/TipTap document
-// tree by node id (#414: the single canonical copy, formerly forked into mcp).
-// Consumed by `@docmost/mcp` (patch/insert/delete node, table tools, outline).
-export {
-  blockPlainText,
-  buildOutline,
-  getNodeByRef,
-  replaceNodeById,
-  deleteNodeById,
-  sanitizeForYjs,
-  findUnstorableAttr,
-  insertNodeRelative,
-  readTable,
-  insertTableRow,
-  deleteTableRow,
-  updateTableCell,
-  assertUnambiguousMatch,
-} from "./node-ops.js";
-export type { OutlineEntry } from "./node-ops.js";
-
-// Normalize a ProseMirror node arg that the model may have serialized as a JSON
-// string (#414: single copy shared by mcp and the CommonJS server app).
-export { parseNodeArg } from "./parse-node-arg.js";
-
-// Inline-footnote authoring convention (#414: single copy, formerly the mcp
-// `footnote-authoring.ts` fork), shared with the importer's `assembleFootnotes`.
-export {
-  footnoteContentKey,
-  makeFootnoteDefinition,
-  generateFootnoteId,
-} from "./footnote.js";

packages/prosemirror-markdown/src/lib/node-ops.ts

@@ -14,8 +14,6 @@
  * `content`, non-object nodes, and absent `attrs` are tolerated.
  */
 
-import { stripInlineMarkdown } from "./text-normalize.js";
-
 /** Deep-clone a JSON-serializable value without mutating the original. */
 function clone<T>(value: T): T {
   if (typeof structuredClone === "function") {
@@ -99,15 +97,12 @@ export function buildOutline(doc: any): OutlineEntry[] {
     const entry: OutlineEntry = {
       index: i,
       type,
-      id:
-        isObject(block) && isObject(block.attrs)
-          ? (block.attrs.id ?? null)
-          : null,
+      id: isObject(block) && isObject(block.attrs) ? block.attrs.id ?? null : null,
       firstText: truncate(blockPlainText(block), 100),
     };
 
     if (type === "heading") {
-      entry.level = isObject(block.attrs) ? (block.attrs.level ?? null) : null;
+      entry.level = isObject(block.attrs) ? block.attrs.level ?? null : null;
     } else if (type === "table") {
       const headerRow = block.content?.[0]?.content ?? [];
       entry.rows = block.content?.length ?? 0;
@@ -252,33 +247,6 @@ export function deleteNodeById(
   return { doc: out, deleted };
 }
 
-/**
- * Throw a clear, model-actionable error when a node-id write op did NOT match
- * exactly one node (#159). `count === 0` -> "no node found"; `count > 1` ->
- * "ambiguous, refused" — Docmost duplicates block ids on copy/paste, so a write
- * by id could clobber/remove EVERY duplicate. The caller skips the write for any
- * `count !== 1` (the transform returns null), so this only REPORTS; nothing was
- * changed. No-op for the unambiguous single-match case.
- */
-export function assertUnambiguousMatch(
-  op: "patch_node" | "delete_node",
-  verb: "replace" | "delete",
-  count: number,
-  nodeId: string,
-  pageId: string,
-): void {
-  if (count === 0) {
-    throw new Error(
-      `${op}: no node with id "${nodeId}" found on page ${pageId}`,
-    );
-  }
-  if (count > 1) {
-    throw new Error(
-      `${op}: id "${nodeId}" is ambiguous — ${count} nodes on page ${pageId} share it (block ids are duplicated on copy/paste). Refusing to ${verb} all of them; nothing was changed. Re-target with a more specific anchor.`,
-    );
-  }
-}
-
 /**
  * Deep-clone `doc` and strip every node/mark attribute whose value is strictly
  * `undefined`, so the result is safe to hand to Yjs (which throws an opaque
@@ -396,31 +364,6 @@ const REQUIRED_CONTAINER: Record<string, string> = {
   tableHeader: "tableRow",
 };
 
-/**
- * Find the index of the first TOP-LEVEL block whose plain text includes the
- * anchor, with a markdown-stripping FALLBACK. Returns -1 when none matches.
- *
- * Two passes preserve "exact wins globally":
- *  - Pass 1: first block containing the verbatim `anchorText`.
- *  - Pass 2 (only if pass 1 found nothing): first block containing the
- *    markdown-stripped anchor, when stripping actually changed it.
- */
-function findAnchorTextIndex(content: any[], anchorText: string): number {
-  if (!Array.isArray(content)) return -1;
-  // Pass 1: exact.
-  for (let i = 0; i < content.length; i++) {
-    if (blockPlainText(content[i]).includes(anchorText)) return i;
-  }
-  // Pass 2: markdown-stripped fallback.
-  const a = stripInlineMarkdown(anchorText);
-  if (a !== anchorText && a.length > 0) {
-    for (let i = 0; i < content.length; i++) {
-      if (blockPlainText(content[i]).includes(a)) return i;
-    }
-  }
-  return -1;
-}
-
 /**
  * Locate an anchor and return its ancestor chain (from `doc` down to and
  * including the matched node). Each chain entry is `{ node, index }` where
@@ -456,14 +399,14 @@ function findAnchorChain(
   }
 
   // By text: only top-level blocks are scanned (same rule as the JSON path).
-  // Exact match wins; a markdown-stripped fallback is tried only on a miss.
   if (opts.anchorText != null && Array.isArray(doc.content)) {
-    const i = findAnchorTextIndex(doc.content, opts.anchorText);
-    if (i !== -1) {
-      return [
-        { node: doc, index: -1 },
-        { node: doc.content[i], index: i },
-      ];
+    for (let i = 0; i < doc.content.length; i++) {
+      if (blockPlainText(doc.content[i]).includes(opts.anchorText)) {
+        return [
+          { node: doc, index: -1 },
+          { node: doc.content[i], index: i },
+        ];
+      }
     }
   }
 
@@ -597,13 +540,13 @@ export function insertNodeRelative(
     return { doc: out, inserted };
   }
 
-  // Resolve by text: only top-level doc.content blocks are scanned. Exact
-  // match wins; a markdown-stripped fallback is tried only on a miss.
+  // Resolve by text: only top-level doc.content blocks are scanned.
   if (opts.anchorText != null && isObject(out) && Array.isArray(out.content)) {
-    const i = findAnchorTextIndex(out.content, opts.anchorText);
-    if (i !== -1) {
-      out.content.splice(i + offset, 0, fresh);
-      return { doc: out, inserted: true };
+    for (let i = 0; i < out.content.length; i++) {
+      if (blockPlainText(out.content[i]).includes(opts.anchorText)) {
+        out.content.splice(i + offset, 0, fresh);
+        return { doc: out, inserted: true };
+      }
     }
   }
 
@@ -674,8 +617,7 @@ function locateTable(
   if (!isObject(rootClone)) return null;
 
   // "#<n>": index into the top-level content array; must be a table.
-  const indexMatch =
-    typeof tableRef === "string" ? tableRef.match(/^#(\d+)$/) : null;
+  const indexMatch = typeof tableRef === "string" ? tableRef.match(/^#(\d+)$/) : null;
   if (indexMatch) {
     const index = Number(indexMatch[1]);
     const block = Array.isArray(rootClone.content)
@@ -775,7 +717,7 @@ export function readTable(
         : undefined;
       const id =
         isObject(firstPara) && isObject(firstPara.attrs)
-          ? (firstPara.attrs.id ?? null)
+          ? firstPara.attrs.id ?? null
           : null;
       rowIds.push(id);
     }
@@ -809,17 +751,14 @@ export function insertTableRow(
   if (!Array.isArray(table.content)) table.content = [];
   const rows = table.content.length;
   const headerRow = table.content[0];
-  const headerCells = Array.isArray(headerRow?.content)
-    ? headerRow.content
-    : [];
+  const headerCells = Array.isArray(headerRow?.content) ? headerRow.content : [];
 
   // Column count is the WIDEST existing row, so the guard below stays
   // meaningful for ragged tables and the new row matches the table's width.
   // Fall back to the supplied cell count only when the table has no rows.
   let colCount = 0;
   for (const r of table.content) {
-    if (isObject(r) && Array.isArray(r.content))
-      colCount = Math.max(colCount, r.content.length);
+    if (isObject(r) && Array.isArray(r.content)) colCount = Math.max(colCount, r.content.length);
   }
   if (colCount === 0) colCount = Array.isArray(cells) ? cells.length : 0;
 
@@ -832,10 +771,7 @@ export function insertTableRow(
   // Resolve the landing index up front so the cell-type decision and the splice
   // below agree: a valid integer in [0, rows] splices there, else we append.
   const landingIndex =
-    typeof index === "number" &&
-    Number.isInteger(index) &&
-    index >= 0 &&
-    index <= rows
+    typeof index === "number" && Number.isInteger(index) && index >= 0 && index <= rows
       ? index
       : rows;
 
@@ -854,8 +790,7 @@ export function insertTableRow(
     // A row landing at index 0 becomes the new header row, so inherit the
     // current header cell's type per column (Docmost uses "tableHeader" there);
     // every other position is a plain data cell.
-    const cellType =
-      landingIndex === 0 ? (headerCells[i]?.type ?? "tableCell") : "tableCell";
+    const cellType = landingIndex === 0 ? headerCells[i]?.type ?? "tableCell" : "tableCell";
     newCells.push({
       type: cellType,
       attrs,
@@ -927,10 +862,9 @@ export function updateTableCell(
   const rowNodes = Array.isArray(table.content) ? table.content : [];
   const rows = rowNodes.length;
   const rowNode = rowNodes[row];
-  const cols =
-    isObject(rowNode) && Array.isArray(rowNode.content)
-      ? rowNode.content.length
-      : 0;
+  const cols = isObject(rowNode) && Array.isArray(rowNode.content)
+    ? rowNode.content.length
+    : 0;
 
   if (
     !Number.isInteger(row) ||

packages/prosemirror-markdown/src/lib/text-normalize.ts

@@ -1,99 +0,0 @@
-/**
- * Locator normalization: strip inline markdown wrappers and trailing
- * decoration from a LOCATOR string so a find/anchor that the model wrote with
- * markdown (or a stray emoji) can still match the document's plain text.
- *
- * This is used ONLY as a fallback for LOCATING (after an exact match fails);
- * it is never applied to replacement text or inserted node content, so no
- * formatting is ever lost.
- *
- * Scope note (#414): this package-local copy exists so `node-ops.ts` — which
- * lives here now (the single canonical copy) — can resolve its markdown-tolerant
- * anchor fallback without a circular dependency back on `@docmost/mcp`. It
- * intentionally carries ONLY `stripInlineMarkdown` (the primitive `node-ops`
- * needs); the mcp-side `text-normalize.ts` (which additionally serves
- * `json-edit.ts` via `stripBalancedWrappers`) is the subject of a separate
- * dedup task and is left untouched here.
- */
-
-/** Maximum unwrap passes, so pathological/nested input cannot loop forever. */
-const MAX_PASSES = 8;
-
-/**
- * Inline emphasis/code/strikethrough wrappers, strong BEFORE emphasis so
- * `**x**` collapses to `x` rather than leaving a stray `*x*`. Each pattern is
- * non-greedy and capture group 1 is the inner text. Applied repeatedly until
- * the string stops changing (nested wrappers like `**_x_**`).
- */
-const WRAPPER_PATTERNS: RegExp[] = [
-  /\*\*([^*]+?)\*\*/g, // **x**
-  /__([^_]+?)__/g, // __x__
-  /~~([^~]+?)~~/g, // ~~x~~
-  /\*([^*]+?)\*/g, // *x*
-  /_([^_]+?)_/g, // _x_
-  /``([^`]+?)``/g, // ``x``
-  /`([^`]+?)`/g, // `x`
-];
-
-/** Links/images -> their visible text. `!?` covers both `[t](u)` and `![a](s)`. */
-const LINK_IMAGE_RE = /!?\[([^\]]*)\]\([^)]*\)/g;
-
-/**
- * Apply the two balanced/link passes: first collapse links/images to their
- * visible text, then collapse balanced inline wrappers repeatedly until stable.
- * Does NOT trim decoration, does NOT guard against an empty result — it returns
- * exactly the transformed string.
- */
-function stripWrappersAndLinks(s: string): string {
-  // 1. Links/images -> their visible text.
-  let out = s.replace(LINK_IMAGE_RE, "$1");
-
-  // 2. Strip balanced wrappers, repeating until the string is stable so nested
-  //    wrappers (`**_x_**`) and adjacent runs both collapse.
-  for (let pass = 0; pass < MAX_PASSES; pass++) {
-    const before = out;
-    for (const re of WRAPPER_PATTERNS) {
-      out = out.replace(re, "$1");
-    }
-    if (out === before) break;
-  }
-  return out;
-}
-
-/**
- * Conservatively strip inline markdown from a locator string.
- *
- * Deterministic, order-fixed steps:
- *  1. Links/images: `[text](url)` -> `text`, `![alt](src)` -> `alt`.
- *  2. Balanced inline wrappers (strong before emphasis, code, strikethrough),
- *     applied repeatedly until stable for nested cases.
- *  3. Trim leading/trailing decoration only: whitespace, leftover marker chars
- *     (`* _ ~ \``) and emoji. Letters/digits and sentence punctuation (`.`/`,`
- *     etc.) are NEVER trimmed.
- *
- * If the result is empty (e.g. the input was only markers like `***`), the
- * ORIGINAL string is returned so a locator can never normalize down to "" and
- * match everything.
- */
-export function stripInlineMarkdown(s: string): string {
-  if (typeof s !== "string" || s.length === 0) return s;
-
-  // 1 + 2. Shared link/image and balanced-wrapper passes.
-  let out = stripWrappersAndLinks(s);
-
-  // 3. Trim leading/trailing decoration: whitespace, leftover markdown markers,
-  //    and emoji (Extended_Pictographic plus the VS16 / ZWJ joiners, plus the
-  //    regional-indicator range U+1F1E6–U+1F1FF for flag emoji, which are NOT
-  //    Extended_Pictographic). The `u` flag enables the Unicode property escape.
-  //    Anchored runs only — interior text and sentence punctuation are untouched.
-  const DECORATION =
-    "[\\s*_~\\x60\\p{Extended_Pictographic}\\u{1F1E6}-\\u{1F1FF}\\u{FE0F}\\u{200D}]+";
-  out = out
-    .replace(new RegExp("^" + DECORATION, "u"), "")
-    .replace(new RegExp(DECORATION + "$", "u"), "");
-
-  // 4. Never normalize a locator down to nothing.
-  if (out.length === 0) return s;
-
-  return out;
-}

pnpm-lock.yaml

@@ -1041,6 +1041,9 @@ importers:
       marked:
         specifier: ^17.0.1
         version: 17.0.5
+      pako:
+        specifier: ^2.0.3
+        version: 2.0.3
       re2:
         specifier: ^1.21.0
         version: 1.25.0