Writing pages

A workspace holds two kinds of human-readable pages: Markdown and HTML. Both
render in this browser. You pick based on how much control you need.

Markdown

Markdown is the default for anything that is mostly words: notes, documentation,
meeting summaries, a personal wiki, a knowledge base like this one.

• Headings, lists, tables, links, code blocks, and quotes all render cleanly.
• It stays readable as raw text, so it travels well and diffs nicely.
• The agent can update one section without disturbing the rest.

This very page is a Markdown file. So are its neighbors in the
knowledge-base/ folder.

When to reach for Markdown

• Notes you will keep adding to.
• Reference docs and runbooks.
• Anything you might later want to grep, quote, or feed back to the agent.

HTML

HTML pages are for when you want a designed result: a landing page, a one-pager,
a styled report, a menu. The agent writes self-contained HTML with its own CSS,
and it renders full-bleed when you open it directly.

HTML pages can run their own scripts, but they are sandboxed: a page can draw
charts and react to clicks, but it cannot reach your Caipi session or other
files. That makes it safe to open a public page someone else's agent wrote.

When to reach for HTML

• A page you want to look a specific way.
• Something a non-technical person will read as a finished artifact.
• A small self-contained widget.

Rule of thumb

If the content is the point, write Markdown. If the presentation is the point,
write HTML. If the page needs to react to a live dataset, you want an app, which
the next page covers.