Separate contiguous paragraphs in HTML passed to the CLI

[robzolkos]

Problem

Basecamp rich text relies on explicit separator nodes for paragraph spacing, not CSS margins on <p>. So contiguous <p>A</p><p>B</p> renders squished — the paragraphs run together and don't look separated. This is common with agent/LLM output, which frequently emits contiguous <p> blocks.

The Markdown pipeline already handles this: a blank line between paragraphs inserts a separator (via TrixBreak). But the raw-HTML passthrough in MarkdownToHTML sent HTML through verbatim, so contiguous <p> from HTML input rendered squished.

Context: internal card "Missing new lines to separate paragraphs in rich text."

Before (main)	After (this PR)
Contiguous <p> renders squished	<br> separators inserted

Change

MarkdownToHTML's HTML-passthrough branch now runs insertParagraphSeparators, which inserts a bare <br> between directly adjacent, non-empty <p> blocks. This brings the raw-HTML path into line with what the Markdown path already does.

Properties:

• Idempotent — a boundary that already carries a separator (a bare <br>, or an empty <p><br></p> / <p></p> on either side) is left untouched, so already-separated content, including Basecamp editor output, is a no-op.
• Byte-preserving apart from the inserted separators.
• Conservative scope — only directly adjacent <p> pairs are separated. Anything between them (a heading, list, or attachment) already provides its own break and is left alone.

Why client-side, and why a bare <br>

Basecamp's editor stores a single-Enter break as contiguous <p>A</p><p>B</p> and a double-Enter break as <p>A</p><p><br></p><p>B</p>. Contiguous <p> is therefore a legitimate, intentional editor state, byte-identical to what an API client sends when it means "separated" — so the server can't safely normalize it without corrupting editor content and breaking round-trips. The CLI can: its input is never editor-authored, and its edit loop already collapses the tight/separated distinction, so treating contiguous HTML paragraphs as separate is consistent with how the CLI already behaves everywhere else.

A bare <br> is used to match what the Markdown pipeline already emits. It renders identically to the editor's canonical <p><br></p> (verified live).

Verification

• New tests in richtext_test.go: behavior + 14 edge cases, idempotency (f(f(x)) == f(x)), and convergence with the Markdown path.
• richtext, commands, output, presenter, tui packages pass; gofmt / go vet / golangci-lint clean; check-skill-drift passes (no CLI surface change).
• Live main-vs-PR comparison against a test project, same input <p>one</p><p>two</p><p>three</p>:
  • main build stores <p>one</p><p>two</p><p>three</p> → renders squished
  • this build stores <p>one</p><br><p>two</p><br><p>three</p> → renders separated

---

Summary by cubic

Ensures HTML passed to the CLI renders separated paragraphs by inserting "<br>" between directly adjacent non-empty <p> tags, matching Markdown behavior. Now treats <p>&nbsp;</p> (and  / ) as empty separator paragraphs.

• Bug Fixes
  • MarkdownToHTML now runs insertParagraphSeparators for HTML input.
  • Inserts "<br>" between adjacent non-empty <p> blocks; preserves existing separators (<br>, empty <p>, or <p>&nbsp;</p>/ / ).
  • Idempotent and byte-preserving; tests added for behavior,   variants, and Markdown-path convergence.

^{Written for commit 8539d0d. Summary will update on new commits.}

[Copilot]

Pull request overview

This PR fixes “squished” rendering when users pass raw HTML with contiguous paragraph blocks into the CLI by normalizing adjacent non-empty <p>...</p><p>...</p> boundaries to include a separator (<br>), aligning the HTML passthrough path with the existing Markdown→Trix behavior.

Tip

If you aren't ready for review, convert to a draft PR.
Click "Convert to draft" or run gh pr ready --undo.
Click "Ready for review" or run gh pr ready to reengage.

Changes:

• Route MarkdownToHTML’s HTML-detection branch through insertParagraphSeparators instead of returning the input verbatim.
• Add paragraph-boundary normalization that inserts <br> only between directly adjacent, non-empty paragraph blocks (whitespace-only gaps preserved; existing separators left untouched).
• Add unit tests covering expected behavior, idempotency, and convergence with the Markdown path.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
internal/richtext/richtext.go	Normalizes raw HTML paragraph adjacency by inserting <br> separators and adds helper for detecting “empty” paragraphs.
internal/richtext/richtext_test.go	Adds test coverage for separator insertion behavior, idempotency, and equivalence with Markdown paragraph breaks.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[cubic-dev-ai]

2 issues found across 2 files

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>

[robzolkos]

Addressed both review findings in 8539d0d:

1. P2 — nbsp separator paragraphs: `isEmptyParagraph` now strips ` `, ` `, and ` `/` ` before the emptiness check, so `

    

   ` separator lines are recognized as empty and left untouched (no extra `
   `). Added tests for all three entity forms.
2. P3 — doc comment: updated `MarkdownToHTML`'s doc comment to reflect that HTML input now gets `
   ` separators inserted between directly adjacent paragraphs, rather than being returned unchanged.

Tests, `gofmt`, `go vet`, and `golangci-lint` all green.

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated no new comments.