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 +1,60 @@
+# Paragraph Deletion via Tracked Changes (WB)
+
+When a reviewer instructs you to delete an entire clause/paragraph, use **paragraph-level deletion** — not just deleting the text but marking the entire paragraph as removed in tracked changes.
+
+## Two-Part Deletion
+
+### Part 1: Paragraph Mark Deletion
+
+Add a `w:del` element inside the paragraph's `w:pPr/w:rPr`:
+
+```xml
+<w:pPr>
+  <w:rPr>
+    <w:del w:id="7777" w:author="WB" w:date="2026-06-26T00:00:00Z"/>
+  </w:rPr>
+</w:pPr>
+```
+
+This marks the paragraph marker (¶) as deleted, so the paragraph doesn't leave an empty line.
+
+### Part 2: Content Deletion
+
+Wrap every text run in the paragraph inside `w:del` elements, converting `w:t` to `w:delText`:
+
+```python
+for child in list(paragraph):
+    tag = child.tag.split('}')[-1]
+    if tag in ('r', 'ins'):
+        paragraph.remove(child)
+        del_elem = etree.SubElement(paragraph, f'{{{W}}}del')
+        del_elem.set(f'{{{W}}}id', '7777')
+        del_elem.set(f'{{{W}}}author', 'WB')
+        del_elem.set(f'{{{W}}}date', '2026-06-26T00:00:00Z')
+        
+        target_runs = child.findall(f'{{{W}}}r') if tag == 'ins' else [child]
+        for r in target_runs:
+            t = r.find(f'{{{W}}}t')
+            if t is not None:
+                r.remove(t)
+                dt = etree.SubElement(r, f'{{{W}}}delText')
+                dt.set('{http://www.w3.org/XML/1998/namespace}space', 'preserve')
+                dt.text = t.text
+            r.set(f'{{{W}}}rsidDel', new_rsid())
+            del_elem.append(r)
+```
+
+### Key Points
+
+- **Same del id** for both the paragraph mark and content dels (e.g., `7777`) — they're part of the same deletion operation
+- **Handle existing INS runs**: If the paragraph has runs inside `w:ins` (from previous revisions), extract them and wrap in `w:del` too
+- **Leave existing DEL runs untouched** — they're already deleted
+- **Copy rPr**: If the original run had `rPr`, copy it into the del run so the strikethrough text renders with correct font/size
+- **Use unique ids**: Pick an id that doesn't collide with existing del/ins ids in the document. Check `max(del_ids) + 1000` if unsure
+
+### Verification
+
+After deletion, render with OnlyOffice and check:
+1. The deleted paragraph appears with strikethrough in markup view
+2. Accepting all revisions removes the paragraph entirely (no empty line)
+3. The paragraph mark (¶) is also deleted — no gap between surrounding paragraphs