[plimeor]

Requested by Plimeor · project thread

Before: three nested empty items were written - - - . An item holding a paragraph -- was written - --. Both were written with no diagnostic, and both read back as a thematic break. Settle's random key test found these.

After: the item's content starts on the line after its bullet, so these are written -\n - - and -\n --. They read back as the same items. This applies to the canonical writer and to an edited item written from the source. Parsed documents such as -\n -- and *\n ** now round-trip exactly.

How:

• starts_below_bullet checks the item's first written line. It moves the content to the next line when that line is padding, or when, after a - or * bullet, the line holds only that char, spaces and tabs.
• Validation's first_item_starts_below_its_bullet uses the same check on item_content_first_line, which follows only the first-child chain. Validation and the writer agree, and validation stays linear.
• The source writer renders such an edited item canonically.
• The validation spec's wording for lists the serializer starts below their bullet now covers this case.
• The plan is archived at docs/archive/2026-10-18-item-first-line-break/.

Checks run locally: every CI command in CLAUDE.md passes. The review pass compared validation's answer with the writer's actual output over 131k random nested-list shapes and found no disagreement.

---

Generated by Claude Code