How Krina handles formatting and rich text in comments

When Krina analyzes a mention, it preserves the text content and most formatting to help the AI understand context. Here’s what gets preserved and what doesn’t.

What formatting is preserved

Bold, italic, and underline

Bold, italic, and underlined text are preserved in the analysis. Krina’s AI reads these as emphasis signals and factors them into classification.

Example:

@Alice Please **URGENT** review the security audit before **Friday EOD**.

Krina recognizes the bold formatting around “URGENT” and “Friday EOD” and weights them heavily in determining urgency.

Code blocks and inline code

Code blocks, code snippets, and inline code are fully preserved. Krina’s AI is trained to understand code context and can classify mentions that reference code differently.

Example:

@Bob The function `validateUser()` is throwing errors.
Here's the stack trace:

TypeError: Cannot read property ‘id’ of null at validateUser (auth.ts:42:15)

Fix this before release.

Krina preserves the code snippet and the stack trace, understanding that this is a critical bug fix request.

Lists and bullet points

Ordered and unordered lists are preserved. This helps Krina understand multi-step requests.

Example:

@Team Please complete the following by end of week:
- Update documentation
- Review pull requests
- Deploy to staging

Krina recognizes this as a multi-item action list and factors in the scope.

Quotes and citations

Block quotes and cited text are preserved, helping Krina understand context when you’re referencing previous comments or requirements.

Links

Hyperlinks and URLs are preserved exactly as they appear. Krina reads the surrounding context to understand what you’re referencing.

What gets normalized

Whitespace and line breaks

Extra spaces and line breaks are normalized to single spaces in some contexts, but the overall structure is preserved so Krina can still understand paragraphs and structure.

Emojis

Emoji characters are preserved as-is. While not “formatting” per se, Krina recognizes emoji patterns like:

• ✅ — Often used to indicate completion or approval
• 🚨 — Often indicates urgency or an alert
• 🎉 — Often indicates success or celebration
• ⏰ — Often indicates a deadline

Example:

@Alice ✅ LGTM! Ship it! 🚀

The emoji reinforces the classification as APPROVE/NOTE.

What’s stripped or not analyzed

Mentions of users (but not the mention itself)

If you reference someone by their display name rather than @mentioning them, that text is preserved but Krina doesn’t create an action item for that person. Only @mentions trigger Krina’s classification.

Jira/Confluence metadata

Jira labels, issue priority, epic names, and other Jira/Confluence system metadata are NOT used in classification — only the human-written comment text matters.

File attachments

Krina doesn’t analyze file content (PDFs, images, documents). It only reads the filename if referenced in the comment text.

Why formatting matters for classification

Formatting helps Krina’s AI distinguish between:

• A casual comment: “hey can you look at this?”
• An urgent request: “@Alice Please review this ASAP. It’s blocking release.”

The bolding signals emphasis, which Krina factors into Urgency. Without that visual context, everything would be treated as equal priority.

Tips for better classification through formatting

1. Use bold for the key ask: @Bob **Please approve** the design changes. → Clearer that approval is needed
2. Highlight deadlines: @Team **Due Friday COB** → Krina catches the deadline better
3. Use code blocks for bug reports: Include full stack traces in code blocks so Krina understands the scope
4. Use lists for multi-step requests: Numbered lists make it clear how many things you’re asking for
5. Use emoji for tone: ✅ for done, 🚨 for urgent, ⏰ for deadline-driven

The more you format, the better Krina understands.

Searching for formatted content

When you search your Krina inbox, the text content is fully searchable regardless of formatting. You can search for “security” and find mentions with bold **security**, italic security, or plain security in your comments.