Arabic PDF Generation (RTL reportlab)

reportlab has no text engine, so an AI agent's first Arabic PDF almost always comes out wrong: letters stand apart instead of joining, each line reads back to front, multi-line paragraphs swap their lines, and English names like iPhone 15 jump out of the sentence. The text extracts fine, so the agent thinks it worked.

This skill gives your agent a tested builder that shapes, reorders and wraps Arabic correctly, line by line.

What it handles

• Shaping and bidi: letters join, words read right to left, mixed English terms and numbers stay in place.
• Line wrapping done right: text is wrapped first and reordered per line, so long paragraphs keep their line order.
• Brackets and mixed text: brackets and numbers never end up on the wrong side.
• RTL tables: columns given in reading order, first column on the right.
• Fonts: finds an Arabic-capable font automatically, or uses yours (Amiri, Noto Naskh…).
• Images: draw_text() writes Arabic labels into Pillow images, using libraqm when available.
• Visual check: preview() renders every page to PNG so the agent can look before delivering.

Works for Arabic, Persian, Urdu and Hebrew. Free and open source (MIT).