Telegram Rich Message Builder

Convert raw message copy into a valid, ready-to-use Telegram `InputRichMessage` payload that the engineering team can pass directly to the backend sendRich method — no manual edits required.

This skill is the product → engineering handoff for Telegram Rich Messages. The PM writes intent and copy; this skill produces the exact structured payload.

Output language rule

• Skill instructions and the "notes for engineering" are in English.
• The message body keeps the original language of the PM's copy. If the copy is Persian, the rendered text stays Persian and is_rtl defaults to true. Never translate the message content.

Inputs to gather

If not already provided, ask the PM for:

1. The raw copy — the text/message they want to send.
2. Structure intent (optional) — e.g. "a title, two bullet points, and a quote at the end". If absent, infer a sensible structure from the copy.
3. Format preference (optional) — markdown (default) or html. Default to markdown because it is readable; switch to html if the copy contains many Markdown-special characters or is assembled from dynamic values where escaping < > & is simpler.
4. Direction (optional) — default is_rtl: true for Persian/Arabic copy; false for purely LTR copy.

What to produce

Always output exactly one InputRichMessage JSON block. This is the object the backend places in rich_message:

{
  "markdown": "<the formatted content>",
  "is_rtl": true
}

or, when HTML is chosen:

{
  "html": "<the formatted content>",
  "is_rtl": true
}

Exactly one of markdown or html — never both, never neither. The backend sendRich method throws otherwise.

Then add a short Notes for engineering block showing the call form:

await telegramSendMessageRepository.sendRich(chatId, { markdown /* or html */: "..." }, { isRtl: true });

Formatting reference — Rich Markdown

Map PM intent to Rich Markdown syntax. This differs from Telegram's legacy `parse_mode` Markdown — use the new syntax:

Formatting reference — Rich HTML (when chosen)

<b>, <i>, <u>, <s>, <code>, <mark>, <tg-spoiler>, <h1>…<h6>, <p>, <pre>, <blockquote>, <a href="…">, lists via <ul>/<ol>/<li>. Escape <, >, & in literal text as <, >, &.

Limits to respect (warn the PM if exceeded)

• ≤ 32768 characters total.
• ≤ 500 blocks (each list item, table row, quote line counts).
• ≤ 16 nesting levels.
• ≤ 50 media attachments; ≤ 20 table columns.

If the copy is over a limit, flag it and propose trimming or splitting into multiple messages.

Media caveat

If the message includes a media block (photo/video/audio), warn: the bot must have media-send permission in the target chat (relevant for groups/channels). Media blocks accept only HTTP/HTTPS URLs and must be their own block.

Constraints

• Output must be a single valid InputRichMessage block — directly usable, no placeholders.
• Preserve the original copy's language and meaning; only add structure/formatting.
• Escape correctly for the chosen format so the payload renders as intended.
• Default is_rtl: true unless the copy is clearly LTR-only.
• Do not invent links, IDs, or data that the PM did not provide.
• Prefer the simplest formatting that conveys the intent — do not over-decorate.

Example

PM raw copy (Persian):

گزارش روزانه. وضعیت سرور سالم است. تعداد درخواست‌ها ۱۲۳۴۰. خلاصه: امروز بدون مشکل.

Output — InputRichMessage:

{
  "markdown": "# گزارش روزانه\n\n- وضعیت سرور: **سالم**\n- تعداد درخواست‌ها: `12,340`\n\n> خلاصه: امروز بدون مشکل.",
  "is_rtl": true
}

Notes for engineering:

await telegramSendMessageRepository.sendRich(
  chatId,
  { markdown: "# گزارش روزانه\n\n- وضعیت سرور: **سالم**\n- تعداد درخواست‌ها: `12,340`\n\n> خلاصه: امروز بدون مشکل." },
  { isRtl: true },
);