@mkterswingman/5mghost-wonder 0.0.17 → 0.0.19

package/dist/wecom/browser-probe.js

@@ -159,6 +159,11 @@ export async function runBrowserNoExportRead(options) {
                 evidence.push("opendoc-response");
             if (extracted?.text)
                 evidence.push("initial-attributed-text");
+            if ((extracted?.tables.length ?? 0) > 0)
+                evidence.push("table-control-characters");
+            if ((extracted?.tables ?? []).some((table) => table.rows.some((row) => row.some((cell) => (cell.colSpan ?? 1) > 1)))) {
+                evidence.push("horizontal-colspan-inference");
+            }
             if (extracted?.structureHints)
                 evidence.push("structure-hints");
             const images = options.saveDir && extracted?.imageUrls
@@ -182,6 +187,7 @@ export async function runBrowserNoExportRead(options) {
                 textLength: extracted?.text.length,
                 imageUrls: extracted?.imageUrls,
                 images,
+                tables: extracted?.tables,
                 structureHints: extracted?.structureHints,
                 extraction: extracted?.text ? "opendoc-initial-attributed-text" : "none",
                 evidence,
@@ -189,7 +195,7 @@ export async function runBrowserNoExportRead(options) {
                     "No readable opendoc initialAttributedText was captured.",
                 ],
                 missing: [
-                    "table merge ranges",
+                    "vertical table merge ranges",
                     ...(images.some((image) => image.status === "downloaded") ? [] : ["image original resources"]),
                     "image anchors",
                     "floating vs fixed image classification",
@@ -314,6 +320,7 @@ export function extractTextFromOpendoc(body) {
     if (!text)
         return null;
     const imageUrls = extractImageUrls(decodedText);
+    const tables = extractTablesFromDecodedText(decodedText);
     const structureHints = analyzeOpendocStructure({
         opendocBody: body,
         htmlData: typeof payload.htmlData === "string" ? payload.htmlData : "",
@@ -325,15 +332,55 @@ export function extractTextFromOpendoc(body) {
         padType: payload.clientVars?.padType ?? payload.padType,
         text,
         imageUrls,
+        tables,
         structureHints,
         warnings: [
             "No-export text is decoded from opendoc initialAttributedText; rich styles are not reconstructed yet.",
+            ...(tables.some((table) => table.rows.some((row) => row.some((cell) => (cell.colSpan ?? 1) > 1)))
+                ? ["Horizontal table colSpan is inferred from row cell counts; vertical rowSpan is not reconstructed yet."]
+                : []),
             ...(imageUrls.length > 0
                 ? ["Image resource URLs were detected, but image anchors and fixed/floating placement are not reconstructed yet."]
                 : []),
         ],
     };
 }
+function extractTablesFromDecodedText(decodedText) {
+    const tables = [];
+    const tablePattern = /\x1a([\s\S]*?)\x1b/g;
+    for (const match of decodedText.matchAll(tablePattern)) {
+        const body = match[1] ?? "";
+        const rows = body
+            .split("\x06")
+            .map((row) => row
+            .split("\x07")
+            .map((cell) => normalizeTableCellText(cell))
+            .filter((text) => text.length > 0)
+            .map((text) => ({ text })))
+            .filter((row) => row.length > 0);
+        inferHorizontalColSpans(rows);
+        if (rows.length > 0)
+            tables.push({ rows });
+    }
+    return tables;
+}
+function inferHorizontalColSpans(rows) {
+    const maxColumnCount = Math.max(0, ...rows.map((row) => row.length));
+    if (maxColumnCount <= 1)
+        return;
+    for (const row of rows) {
+        if (row.length !== 1)
+            continue;
+        row[0].colSpan = maxColumnCount;
+    }
+}
+function normalizeTableCellText(value) {
+    return value
+        .replace(/\r/g, "")
+        .replace(/[\u0000-\u0006\u0008-\u001f\u007f-\u009f]+/g, "")
+        .replace(/[^\S\r\n]+/g, " ")
+        .trim();
+}
 function analyzeOpendocStructure(input) {
     return {
         opendoc: summarizeStructureSignals(input.opendocBody),

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mkterswingman/5mghost-wonder",
-  "version": "0.0.17",
+  "version": "0.0.19",
   "description": "企微文档读取 CLI — WeCom document reader",
   "type": "module",
   "engines": {

package/skills/setup-5mghost-wonder/SKILL.md

@@ -7,9 +7,9 @@ description: Use this skill when the user wants to install or set up wonder, say
 
 ## Skill version
 
-This skill matches **wonder 0.0.17**.
+This skill matches **wonder 0.0.19**.
 
-Once the CLI is installed in Step 1, run `wonder --version`. If the output does not equal `0.0.17`, the CLI on disk has drifted from the skill text loaded in this session. Ask the user to run `/update-5mghost-wonder`, then **start a fresh AI session** (`/exit` and re-enter, or open a new chat) — skill text already loaded into a running session does not refresh after `wonder update`, even though the file on disk has been replaced.
+Once the CLI is installed in Step 1, run `wonder --version`. If the output does not equal `0.0.19`, the CLI on disk has drifted from the skill text loaded in this session. Ask the user to run `/update-5mghost-wonder`, then **start a fresh AI session** (`/exit` and re-enter, or open a new chat) — skill text already loaded into a running session does not refresh after `wonder update`, even though the file on disk has been replaced.
 
 After a successful first install, also remind the user to start a fresh AI session before invoking `/use-5mghost-wonder` for the first time. The skill files were just written to disk; the current session never loaded them.
 

package/skills/update-5mghost-wonder/SKILL.md

@@ -7,7 +7,7 @@ description: Use this skill when the user wants to update or upgrade wonder, say
 
 ## Skill version
 
-This skill matches **wonder 0.0.17**.
+This skill matches **wonder 0.0.19**.
 
 ---
 

package/skills/use-5mghost-wonder/SKILL.md

@@ -10,10 +10,10 @@ the referenced workflow files needed for the current task.
 
 ## Version Gate
 
-This skill matches **wonder 0.0.17**.
+This skill matches **wonder 0.0.19**.
 
 On first use in a session, follow `references/session-init.md`. If the installed
-CLI version differs from `0.0.17`, stop and ask the user to run
+CLI version differs from `0.0.19`, stop and ask the user to run
 `/update-5mghost-wonder`, then start a fresh AI session.
 
 ## Hard Rules

package/skills/use-5mghost-wonder/references/read-no-export-browser.md

@@ -37,10 +37,12 @@ wonder browser read <url> --save /tmp/wonder-browser-read
 ```
 
 This currently attempts text extraction from the browser `opendoc`
-`initialAttributedText`, downloads detected image resources when `--save` is
+`initialAttributedText`, decodes simple table cells from WeCom table control
+characters, infers horizontal `colSpan` when a single-cell row spans the
+table's max row width, downloads detected image resources when `--save` is
 present, and returns safe `structureHints` for `htmlData`,
 `initialAttributedText`, and the full `opendoc` body. Treat it as `partial`
-unless it also proves actual table cells, merge ranges, and image anchors.
+unless it also proves vertical merge ranges and image anchors.
 
 3. If the read is insufficient, run the evidence probe:
 
@@ -61,11 +63,13 @@ Use `--headed` only when login/debug visibility is needed.
    - WebSocket messages if they expose structured document operations
    - screenshots only as a fallback or visual cross-check
 6. For tables, specifically look for:
+   - `tables[].rows[].text` for simple table cell content
+   - `tables[].rows[].colSpan` for inferred horizontal merged cells
    - `structureHints.*.hasTableSignals` and `hasMergeSignals`
    - `structureHints.htmlData.hasTableMarkup`
    - cell coordinates
    - displayed text
-   - merge ranges
+   - vertical merge ranges / `rowSpan`
    - row/column sizes if available
    - fixed cell images
    - floating images with position, size, and anchor
@@ -86,6 +90,14 @@ Return a concise result:
   "mergeRangesAvailable": true,
   "imagesAvailable": true,
   "imageAnchorsAvailable": true,
+  "tables": [
+    {
+      "rows": [
+        [{ "text": "A1B1", "colSpan": 2 }],
+        [{ "text": "A2" }, { "text": "B2" }]
+      ]
+    }
+  ],
   "structureHints": {
     "htmlData": {
       "hasTableMarkup": true,

package/skills/use-5mghost-wonder/references/verification.md

@@ -41,7 +41,8 @@ The goal is lossless structured data. Verify each field independently:
 
 - text
 - table cell coordinates
-- merge ranges
+- horizontal `colSpan`
+- vertical merge ranges / `rowSpan`
 - image original resources or URLs
 - image anchors, positions, and sizes
 - floating vs fixed cell images