memories: lint-markdown's fail input does not cover its companion checks

[d-morrison]

Closes #2503

One entry appended to memories/gha-reusable-workflows.md.

lint-markdown.yml's fail input gates markdownlint alone. Its list-item-splice,
code-block-length, and table-split companions read fail-on-item-splices,
fail-on-long-code-blocks, and fail-on-table-splits, each defaulting to true --- so
a caller written fail: false still fails the build. lint-yaml.yml splits the
same way (fail versus fail-on-long-scripts).

Worth an entry rather than a shrug because the evidence reads as a
contradiction: the caller says warn-only, the check is red, and the two
hypotheses that present themselves --- wrong pin, ignored input --- are both
wrong. The job's conclusion came from a different step, which is the shape
fully-clean.md already describes for a green guard step beside a red job.

Measured on UCD-SERG/shigella#37,
whose caller was added at fail: false in
#33 and went red on six
list-item splices.

scripts/semantic-line-breaks.py --base origin/main reports 0 paragraphs would
change; no non-ASCII punctuation added.

[d-morrison]

Addressed in 4490d988

The finding is correct, and correct about a claim I had no business making
loosely in a memory file. Verified against the @v2 tag consumers actually pin
(the review checked main; both agree here, but @v2 is the one that governs):

check	input	default at @v2
markdownlint	fail	true
list-item merge splices	fail-on-item-splices	true
GFM table splits	fail-on-table-splits	true
fenced-code-block length	fail-on-long-code-blocks	false

The entry now carries that table rather than a summary sentence, so each
default is checkable individually.

The correction also sharpened the entry rather than merely repairing it. My
draft cited lint-yaml as "the same split", implying it bites the same way ---
but fail-on-long-scripts defaults to false there, so fail: false on
lint-yaml genuinely is warn-only. That contrast is now the entry's point:
the split existing tells you nothing, only the defaults do, and they differ
within a single workflow.

The provenance note records that the first draft got this wrong and that the
review caught it.

Posted by Claude Code (AI agent) --- not written by a human.

[github-actions]

Claude finished review — View run

---

Both PRs check out and match the described content — #37 is open and its body confirms the exact same correction narrative (defaults table matching the memory entry exactly, including the "review caught it" provenance), #33 is merged and matches the migration description. The fully-clean.md cross-reference resolves.

No content defects found. This is a memory-only change; the factual error flagged in the prior round (5b318af) is now corrected in the current diff (4490d98/HEAD), and all four lint-markdown defaults plus the lint-yaml comparison are verified accurate against the live Morrison-Lab/gha@v2 workflow files. No ASCII-punctuation or formatting violations, Do/Don't pairing is followed, and the entry structure matches the rest of the file.

This confirms the prior round's verdict, posted at #2504 (comment) (2026-08-28T05:29:53Z): Ready for merge. No new findings in the current diff.

Verdict

Ready for merge — content re-verified against the live Morrison-Lab/gha@v2 workflow files (all four lint-markdown defaults and the lint-yaml comparison confirmed accurate) and against both referenced shigella PRs (#37 open, #33 merged, both matching the described content). No new issues found; the prior round's verdict stands.

Reviewed commit: 77e4112

[github-actions]

💰 Cost: $0.5409 (review) — run