برومبت اصلاح مشاكل العربي RTL في Word وPDF مع أي أداة ذكاء اصطناعي

ARABIC DOCS (.docx/PDF) — CORE RULES
0) If you cannot inspect/edit raw XML, or can only output chat text with no real downloadable file, say so explicitly — never present plain text as if it were a Word/PDF file.
1) Every Arabic paragraph=bidi, every Arabic run=rtl=”1″ (Latin runs: rtl=”0″), every table=bidiVisual. Apply in body, headers, footers, and notes — not just main text.
2) NEVER set jc=”right”/”left” (or text-align) on a bidi/RTL paragraph. jc is direction-relative and flips to the opposite side in real Word even when a preview looks correct. Omit jc entirely on Arabic paragraphs; only use it for center/justify.
3) Keep pPr/rPr/tblPr/sectPr children in OOXML schema order — Word silently drops out-of-order elements (LibreOffice hides this bug).
4) Mixed Arabic/Latin: codes, URLs, dates, brackets = one unsplit LTR run; never reorder the actual words.
5) Native tables only (no Markdown pipes), RTL column order, ≥170 twips cell padding.
6) PDF: never ReportLab/FPDF/jsPDF/matplotlib. Use HTML dir=”rtl” + WeasyPrint/Chromium, LibreOffice conversion, or XeLaTeX.
7) MANDATORY AUDIT before delivery: search every XML part (body/header/footer/notes) for bidi co-occurring with jc=”right”/”left” — delete any match found. Confirm every table has bidiVisual and every Arabic run has rtl=”1″. If you cannot verify this, say so and deliver RTL-correct HTML instead of an unverified file.