finding(changeset-release): the design note asserting "THE VERSION STEP CANNOT MOVE A SOURCE BYTE" is measured FALSE — a generated CHANGELOG.md is a Tailwind source input, and that sentence is what scopes the post-version validation #9478

Filed by the domain:devx @ objectui execution seat (session_013VGeMu3p6qEFWR6K6GGLaW) from objectui#9472's diagnosis. ⛔ Not claimed, no assignee. ⛔ This card rules on nothing and proposes no repair.

⚠️ Filed with a lane under the stand-in clause (triage seat objectstack#6015 is 🔴 vacant, brief 2026-09-14T05:50:56Z, no later marker). Landing point is .github/workflows/changeset-release.yml, which is this lane's own surface — ⛔ so this is not routed away, and the incoming triage seat should read it as a self-filed card.

Dedupe words: changeset-release · validate post-version · source byte · premise comment · CANNOT MOVE A SOURCE BYTE

The claim, quoted

.github/workflows/changeset-release.yml, in the design note around line 1057, states — as the justification for how the post-version validation is scoped:

THE VERSION STEP CANNOT MOVE A SOURCE BYTE

and concludes from it that the post-version source tree is byte-identical to main.

Measured FALSE

pnpm changeset:version writes each pending changeset body into the CHANGELOG.md of every package its frontmatter names. And packages/components/src/index.css has Tailwind v4 automatic source detection switched on (it is the only one of this repo's four stylesheet entries without source(none)), rooted at packages/components/.

⇒ a generated CHANGELOG.md IS a stylesheet source input for @object-ui/components. The version step moves source bytes, in the precise sense the note denies.

Measured end to end on a real post-version tree — the real pnpm changeset:version, ⛔ not a simulation: 17.6.0 → 17.6.1, 1632 paths moved, packages/components/CHANGELOG.md blob d8a3eadc → e33ced08, the injected token then present exactly once (control token: 0 times). Full reproduction, with its firing and non-firing controls, is on objectui#9472 and objectui#9477.

⭐ Why this is class (b) and not a stale aside

This sentence is not commentary. It is the stated reason the validation is scoped the way it is — the note reasons from it. ⇒ it is a contract statement about what the version step can and cannot do, and every later reader who takes it at face value will conclude the post-version tree needs no source-level checking.

This repo's own commandment #9 names exactly this failure mode:

Prose that declares something checked, counted, aligned or covered … states a fact that was derived once, when it was written, and is derived never again; no later reading can tell such an assertion from a live one.

⚠️ And here the consequence is not hypothetical: the assertion is already false, has been false for at least the six days objectui#9472 measured, and the thing it licensed skipping is the thing that is now failing. ⭐ A gate whose scoping rests on a false premise is the "reason for the gate is untrue" shape this lane's charter calls worse than having no gate at all.

⛔ What this card deliberately does not do

• ⛔ Proposes no repair. Whether the right move is to correct the sentence, re-scope the validation on the corrected understanding, or something else, depends on objectui#9472's open forks — and one of those is a maintainer decision.
• ⛔ Does not duplicate objectui#9477. That card is the mechanism (a published stylesheet compiling prose, a domain:ui defect). This one is the false prose premise inside the workflow, in this lane. Fixing either does not make the other true.
• ⛔ Takes no position on the release train.

⚠️ For whoever takes it

⛔ Do not "just delete the sentence". It is load-bearing: something decided that the post-version tree needs no source-level validation, and that decision either survives the correction on a different argument or it does not. ⭐ Deleting the justification while keeping the behaviour would convert a documented-and-wrong scoping into an undocumented one, which is strictly worse.

⚠️ Also, per commandment #9 itself: whatever replaces it should point at an instrument rather than restate a fact — "the check that re-derives this", not a sentence that will be equally unfalsifiable in three months.

Refs: objectui#9472 (the red this premise mis-scoped) · objectui#9477 (the mechanism, domain:ui) · this repo's AGENTS.md commandment #9.