Two more frontmatter splitters remain after #162, both deliberately weaker — record the difference before someone consolidates them

[hotlong]

The finding

Observation, not a defect — filing it for triage rather than acting on it.

The #162 sweep looked for frontmatter-splitting copies across scripts/. It found the two the card named (content-lint.mjs, lib/post-dates.mjs, now consolidated into scripts/lib/frontmatter.mjs) and two more the card did not:

Where	Shape	Fails on a bad fence?	YAML
scripts/lib/frontmatter.mjs (after #162)	startsWith('---\n') + indexOf('\n---', 4)	yes — reports no-opening-fence / no-closing-fence, caller throws	js-yaml, in the caller
scripts/gen-zh-hant.mjs:168 splitFrontmatter(source)	/^---\n[\s\S]*?\n---\n/	no — falls back to { head: '', body: source }	none (head is copied verbatim)
scripts/lib/wechat-html.mjs:43 splitFrontmatter(src)	/^---\n([\s\S]*?)\n---\n?/	no — falls back to { data: {}, body: src }	none — single-line scalars read with /^(\w+):\s*(.+)$/

scripts/new-post.mjs has no copy: it only writes a frontmatter template. scripts/to-wechat.mjs has no copy either; it delegates to lib/wechat-html.mjs.

Why they were left alone

Both differences look deliberate, and #162 was a pure consolidation card that was explicitly not allowed to ship a behaviour change:

• Neither is a gate. gen-zh-hant.mjs must still rewrite a body it cannot read a head from, and the WeChat exporter must still produce HTML — so both tolerate what the lint copy rejects. Routing either through the shared helper would turn a tolerated file into a failing one.
• wechat-html.mjs deliberately avoids js-yaml because it only ever wants title / description / author, and its comment says so: "Minimal frontmatter splitter — single-line scalars only".
• Both require a newline after the closing fence (\n---\n), which the shared helper does not — a file ending exactly at \n--- splits under content-lint.mjs and does not split under these two.

The differences are now recorded in a comment at the bottom of scripts/lib/frontmatter.mjs, so the next reader finds them documented rather than discovering them. This issue exists so the decision is on record too.

What a triage might decide

1. Leave both — they serve different jobs and the difference is documented. Cheapest, and the status quo.
2. Give the helper a tolerant mode (splitFrontmatter(source, { strict: false })) and route all four through it, so there is one fence definition with two failure policies. This is a behaviour change for whichever file today falls in the gap, and needs its own equivalence table across pnpm gen:zh-hant and pnpm wechat.
3. Leave wechat-html.mjs alone permanently — a regex scalar reader is not a frontmatter parser and should not pretend to be one; only gen-zh-hant.mjs is a candidate.

Not urgent either way: nothing is broken today, and the two weak splitters agree with each other.

Found while implementing #162. scripts/gen-zh-hant.mjs was read-only for that round (owned by #194).

Generated by Claude Code