[ChangkeunJ]

Fixes #1628.

Markdoc's formatter only escapes a leading *, # or > in a text node, plus *, _ and ~ directly inside strong, em and s. So when the markdoc or document field saves, a paragraph stored as 2026\. A turning point comes back as 2026. A turning point and is an ordered list on the next parse, \- not a list turns into a bullet, a \*literal\* word turns into emphasis and \{% braces %} turns into an unknown tag that stops the entry from opening again (same symptom as #1428).

Every format call in the package already goes through src/markdoc.js, so this wraps format there. Before formatting it walks the AST and escapes the text that would otherwise be read back as syntax: {% everywhere, * and _ outside strong, em and s (Markdoc already escapes those inside), and a -, +, 1. or 1) marker at the start of a line. Headings are skipped so ## 1. Intro doesn't change, a _ between two word characters is skipped so snake_case doesn't churn, and a * or _ with spaces on both sides can't open or close emphasis so it's skipped too. The original text is put back after formatting, so the caller's AST isn't changed.

One thing to be aware of: existing content with a bare * or _ next to a word will pick up a backslash on its next save. It renders the same, but it is a diff, so I kept the rule as narrow as I could. A literal backslash, backticks and ~~ outside an s node aren't covered; I left those out rather than guess at what the editor should do with them.

The new test in markdoc.test.tsx has the cases from the issue and fails on main. The keystatic test suite, tsc, eslint and prettier pass with the change.

[changeset-bot]

🦋 Changeset detected

Latest commit: f0d42af

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
@keystatic/core	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR