feat: export core Hermes skills

2026-07-15 02:45:56 +00:00
parent a028b63eda
commit 54711fee2a
308 changed files with 41310 additions and 1 deletions
@@ -0,0 +1,93 @@
+# Strip numPr Before Inserting Manual Numbering (2026-07-01)
+
+## Problem
+
+When adding manual numbering (e.g. "第一条  ") as `w:ins` at the beginning of a paragraph that already has `<w:numPr>` (automatic numbering like "%1." decimal format), OnlyOffice renders BOTH:
+- The automatic number: "1."
+- The manual INS text: "第一条"
+
+Result: "1. 第一条  乙方应严格按照..."
+
+## Root Cause
+
+`<w:numPr>` in pPr tells the rendering engine to prepend an auto-generated number. The `w:ins` text is just another run in the paragraph — it doesn't suppress the auto-numbering.
+
+Additionally, `<w:pPrChange>` records the pre-revision pPr state. If pPrChange still contains `<w:numPr>`, some renderers will show the old numbering in markup view.
+
+## Affected Scenarios
+
+1. **反委托代发工资协议 (2026-07-01)**: Original paragraphs P6-P12, P19 had `numId=1` or `numId=3` (decimal "%1." format). After adding "第一条" through "第十一条" as INS, OnlyOffice showed "1. 第一条", "2. 第二条", etc.
+
+2. **Any contract where the original used auto-numbering**: Check `numbering.xml` for active numId references with `numFmt=decimal` or `numFmt=chineseCounting`.
+
+## Fix Pattern
+
+```python
+import zipfile
+from lxml import etree
+
+WNS = '{http://schemas.openxmlformats.org/wordprocessingml/2006/main}'
+
+with zipfile.ZipFile(docx_path, 'r') as z:
+    all_files = {name: z.read(name) for name in z.namelist()}
+
+doc = etree.fromstring(all_files['word/document.xml'])
+body = doc.find(f'{WNS}body')
+paras = body.findall(f'{WNS}p')
+
+# Identify paragraphs where we added manual numbering INS
+# These are paragraphs that have both:
+# 1. A w:ins with author=WB containing "第X条" text
+# 2. A pPr with numPr
+
+for i, p in enumerate(paras):
+    ppr = p.find(f'{WNS}pPr')
+    if ppr is None:
+        continue
+    
+    # Check if this paragraph has our manual numbering INS
+    has_manual_numbering = False
+    for child in p:
+        tag = child.tag.split('}')[-1] if '}' in child.tag else child.tag
+        if tag == 'ins' and child.get(f'{WNS}author') == 'WB':
+            text = ''.join(t.text for t in child.iter(f'{WNS}t') if t.text)
+            if '第' in text and '条' in text:
+                has_manual_numbering = True
+                break
+    
+    if not has_manual_numbering:
+        continue
+    
+    # Strip numPr from pPr
+    num_pr = ppr.find(f'{WNS}numPr')
+    if num_pr is not None:
+        ppr.remove(num_pr)
+        print(f"P{i}: Stripped numPr from pPr")
+    
+    # Strip numPr from pPrChange
+    ppc = ppr.find(f'{WNS}pPrChange')
+    if ppc is not None:
+        inner_ppr = ppc.find(f'{WNS}pPr')
+        if inner_ppr is not None:
+            inner_num = inner_ppr.find(f'{WNS}numPr')
+            if inner_num is not None:
+                inner_ppr.remove(inner_num)
+                print(f"P{i}: Stripped numPr from pPrChange")
+
+# Save back
+all_files['word/document.xml'] = etree.tostring(doc, xml_declaration=True, encoding='UTF-8', standalone=True)
+# Write to temp file then replace (never write to same zip you're reading)
+```
+
+## Verification
+
+After stripping, render with OnlyOffice and confirm:
+1. Markup view shows only the manual numbering (no "1." prefix)
+2. Accepted-revisions view shows clean "第一条" through "第十一条"
+3. python-docx can still open the file without errors
+
+## Edge Cases
+
+- **DEL-only empty paragraphs** (e.g. P8, P9 where all content is w:del): These may still have numPr. If they render a visible "3." or "4." in the gap, strip those too.
+- **Cross-paragraph clauses** (P6+P7 = one clause): P7 may have its own independent numPr even though it's a continuation paragraph. Strip it.
+- **numId=0 (disabled numbering)**: `numId=0` in OOXML means "numbering OFF" — it doesn't render anything. Only strip numPr where `numId > 0` and the corresponding abstractNum has a visible numFmt (decimal, chineseCounting, etc.).