Skip to content

docs(contributing): a repository count in a changeset is a reading at a named commit - #10978

Merged
objectstack-fleet[bot] merged 2 commits into
mainfrom
claude/issue-9649-changeset-count-anchor
Sep 28, 2026
Merged

objectstack-fleet[bot] merged 2 commits into
mainfrom
claude/issue-9649-changeset-count-anchor

Conversation

@objectstack-fleet

Copy link
Copy Markdown
Contributor

Fixes #9649

Clause-②: no — contributor guidance in a non-governed, unpublished file; no contract, accept set or public surface changes

What changes

CONTRIBUTING.md, section ### Changeset Message Guidelines only: one short rule and one good/bad pair in the section's existing example style. Diff: +14 lines, no other section touched.

  • A count over the repository ("N files", "N call sites", "N occurrences") in a changeset is a reading of one tree. It publishes verbatim, nothing re-checks it, and the next file added can falsify it, including a sibling release note that quotes the same token.
  • So it is stated in the past tense as a reading at a named commit, or left out. "This branch" means nothing once the note is in the CHANGELOG, so the example names the commit.
  • It points at AGENTS.md commandment #9 as the general rule and does not restate it.

Route

The card offered (i) a gate, (ii) a convention, (iii) nothing, and ruled on none of them. The claiming seat chose (ii), and that choice is recorded in the claim comment on the card:

  • (i) is not taken. The four-axis clause says new gates default to no unless a maintainer names one, and no naming is on record.
  • (iii) is not taken. It would leave user-visible release-note claims that cannot be revised after publishing.

No checker, script, lint rule or test that polices changeset prose is added.

Premise checks (on origin/main 328abeb)

  • H1, confirmed. The section had a good/bad list and 0 lines about counts, revisions or tense. node scripts/check-governed-queue-guard.mjs --test CONTRIBUTING.md answered NOT GOVERNED. The control --test AGENTS.md answered GOVERNED, exit 3. Root AGENTS.md commandment #9 states the general principle: prose that declares something counted is derived once and never again, and where nothing re-derives it the text must say so. It does not name changesets or the named-commit device, so the new text points at #9 instead of restating it. .changeset/README.md has no message-writing guidance.
  • H2, confirmed. .changeset/6951-tree-view-data-retired.md is still pending on main (git cat-file -e exit 0, with .changeset/config.json as the control, also exit 0). Its figures are headed "What was measured, on this branch's base." The note names no commit, so a CHANGELOG reader cannot resolve "this branch's base". The new guideline asks for the commit to be named.
  • H3, re-derived. The card said "five notes", and that figure had not been re-derived. It is an undercount. The per-changeset list, with a truth reading on main for each entry, is in the os-dev-report comment on the card. No changeset is edited here.

Local verification (head 247cf09)

  • node scripts/check-doc-links.mjs → exit 0, "Links are valid across 17 scan roots."
  • pnpm check:control-bytes → exit 0, "OK (scanned 9250 tracked text file(s); skipped 85 binary)"
  • pnpm check:new-line-citations → exit 0, "VERDICT new-cross-file-line-citations: 0 new citation(s), enforcement report-only"
  • node scripts/check-changeset-presence.mjs → exit 0, "No source or published contract of a released package changed in this range, so no changeset is owed."
  • pnpm check:installed-pin-claims → exit 0, "OK"
  • The tests that read CONTRIBUTING.md are scripts/__tests__/check-installed-spec-pin-claims.test.ts and scripts/__tests__/ci-cd-pipeline-doc.test.ts, per the markdown-test-inputs ledger. The grep hit scripts/__tests__/check-doc-links.test.ts was also run, although it only uses fixtures. They ran as pnpm exec vitest run --maxWorkers=2 under the shared verify lock → "Test Files 3 passed (3), Tests 235 passed (235)", VERDICT command-exit 0.
  • pnpm check:doc-snippets is not owed. Its header lists root pages other than README.md, CONTRIBUTING.md included, as outside its roots. A local invocation exited 2 on unbuilt packages (a prerequisite, not a measurement), so this is NOT MEASURED and was not read as red.

Acceptance notes

  • The section's existing ✅ example "Improve performance of large data grids by 50%" is a measured figure, but it is not a count over the repository. It is left untouched, per the file surface.
  • .changeset/6951-tree-view-data-retired.md also carries a cross-file line address in its "measured" paragraph, the shape AGENTS.md #11 bans. That is an observation only: it is a pending body owned by another card, it is outside this card's file surface, and no one is taking it up.

Attribution: this change was made in session https://claude.ai/code/session_01EBx9rvB7dufCz4at53x35U.


Generated by Claude Code

… a named commit (objectui#9649)

A changeset publishes verbatim into the CHANGELOG, days after it was written,
and nothing re-checks a figure in it. A count over the repository ("N files",
"N call sites") is therefore a reading of one tree presented as a claim about
the released code, and the next file added can falsify it, a sibling release
note quoting the same token included.

The Changeset Message Guidelines section gains one short rule and one good/bad
pair: state such a count in the past tense as a reading at a named commit, or
leave it out. It points at AGENTS.md commandment #9 as the general rule rather
than restating it.

Claude-Session: https://claude.ai/code/session_01EBx9rvB7dufCz4at53x35U
Co-authored-by: Claude <noreply@anthropic.com>
…us (objectui#9649)

"read" is spelled the same in the present and the past, so the pair did not
show the tense difference the rule asks for. "used" / "use" does.

Claude-Session: https://claude.ai/code/session_01EBx9rvB7dufCz4at53x35U
Co-authored-by: Claude <noreply@anthropic.com>
@github-actions github-actions Bot added the documentation Improvements or additions to documentation label Sep 28, 2026
@objectstack-fleet
objectstack-fleet Bot marked this pull request as ready for review September 28, 2026 12:36
@objectstack-fleet
objectstack-fleet Bot added this pull request to the merge queue Sep 28, 2026
Merged via the queue into main with commit a2de9e9 Sep 28, 2026
41 checks passed
@objectstack-fleet
objectstack-fleet Bot deleted the claude/issue-9649-changeset-count-anchor branch September 28, 2026 12:50
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

documentation Improvements or additions to documentation

Projects

None yet

2 participants