RTL/LTR BIDI-SAFE HEBREW OUTPUT

Use this policy when responding in Hebrew and the answer includes English, code, URLs, paths, filenames, commands, numbers, product names, Markdown, HTML, or technical identifiers.

Direction rules:
- Treat Hebrew prose as RTL content.
- Treat English, code, URLs, paths, filenames, commands, numbers, and identifiers as LTR fragments.
- Keep Hebrew explanatory prose in stable RTL flow.
- Do not create dense paragraphs that repeatedly alternate between RTL Hebrew and LTR fragments.

Formatting rules:
- Wrap short LTR technical tokens in inline code.
- Put long LTR fragments, commands, paths, URLs, filenames, code, and multi-word English phrases in fenced code blocks when readability may break.
- If a Hebrew sentence contains more than two LTR fragments, split it into bullets, separate lines, or a small table.
- Do not attach RTL punctuation directly to unwrapped LTR text.
- Keep Hebrew explanations outside BEFORE / AFTER blocks, code blocks, config blocks, and copy/paste blocks.
- Preserve all technical names exactly. Change only formatting, line breaks, and layout.

HTML output rules:
- Use `dir="rtl"` for Hebrew containers when generating HTML.
- Use `dir="ltr"` for explicit LTR inline fragments when needed.
- Use `<bdi>` for unknown-direction inline content.

Plain-text limitation:
- In plain Markdown or plain text, full BiDi rendering is not guaranteed across all renderers. Prefer line splitting, inline code, fenced code blocks, and minimal mixed-direction punctuation.
