fix(bulma-ui): make Theme colorMode switch the scheme under the prefixed builds

[allxsmith]

<Theme colorMode="dark"> did nothing under the prefixed builds. Theme writes data-theme, but Bulma's theme mixin emits [data-#{$class-prefix}theme=…], so bestax-prefixed and bestax-no-helpers-prefixed key their schemes off data-bestax-theme and carry no data-theme hook at all. Dark on a light OS stayed light, and light on a dark OS stayed dark.

• Theme reads the class prefix from ConfigProvider, and under one, colorMode writes data-<prefix>theme alongside data-theme, from scoped and root Themes alike. 'system' removes both, and unmounting or changing colorMode restores each to what it was.
• With no prefix, nothing changes: only data-theme is written, with the same DOM operations as before, and no CSS moved.
• A prefix that can't form an attribute name (md/, say) still gets data-theme, as it did before, instead of throwing.

A new styles test compiles every published sheet and resolves the scheme on <html> for dark, light, system and unmount, scoped and root, on a light and a dark OS. On main every Theme case failed on both prefixed sheets and passed on the unprefixed ones. Chromium agrees: the prefixed sheets ignore data-theme alone and switch with both. Unit tests cover the no-prefix case, both attributes, isRoot, system, restoring on unmount and an invalid prefix.

The colorMode TSDoc moved onto the prop and says both attributes are written; the Theme page and the theming skill's dark mode reference say the same, and note that [data-theme='dark'] custom CSS still matches.

pnpm all passes locally.

[coderabbitai]

Warning

Review limit reached

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

Next included review available in 31 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used the included review currently available.

Learn how review limits work.

Review configuration:

⚙️ Run configuration

• Configuration used: Repository: allxsmith/bestax/.coderabbit.yaml
• Review profile: CHILL
• Plan: Advanced
• Run ID: a0ae2bb6-65a3-4022-a227-8b3198cb4707

📥 Commits

Reviewing files that changed from the base of the PR and between 99426b4 and 81b02df.

📒 Files selected for processing (15)

• bestax-mcp/data/components/Theme.json
• bestax-mcp/data/skills.json
• bulma-ui/src/helpers/CLAUDE.md
• bulma-ui/src/helpers/Theme.tsx
• bulma-ui/src/helpers/__tests__/Theme.styles.test.tsx
• bulma-ui/src/helpers/__tests__/Theme.test.tsx
• docs/docs/api/helpers/theme.md
• docs/docs/guides/features/css-variables.md
• docs/docs/guides/getting-started/optimizing-css.md
• docs/docs/guides/helpers/color.md
• docs/docs/skills/theming.mdx
• skills/README.md
• skills/bestax-theming/SKILL.md
• skills/bestax-theming/examples/dark-mode.tsx
• skills/bestax-theming/references/css-variables.md

• Autopilot · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

Preview Deployment

Preview URL: https://1707c222.bestax.pages.dev

[claude]

Deep review — 0 blocking · 3 advisory

No.	Severity	Area	Finding	Location
1	🔵 Advisory	Correctness	colorMode is documented as "always global", but the prefixed attribute now depends on where Theme sits relative to ConfigProvider — a Theme colorMode mounted above the provider writes only data-theme, and a prefixed sheet still ignores it	bulma-ui/src/helpers/Theme.tsx:1084
2	🔵 Advisory	Robustness	The catch is silent, so a classPrefix that is a legal CSS ident but not a legal attribute name (any non-ASCII — an emoji prefix, say) makes Bulma emit a selector the DOM can never satisfy, and colorMode fails with no diagnostic	bulma-ui/src/helpers/Theme.tsx:1096
3	🔵 Advisory	API	Three other surfaces still describe colorMode as writing data-theme with no prefixed caveat (skills/bestax-theming/SKILL.md:27, docs/docs/guides/features/css-variables.md:236, Theme.stories.tsx:572), and the prop-table parenthetical names data-bestax-theme where a classPrefix="my-" yields data-my-theme. All still true, just incomplete — and SKILL.md is what an agent reads before the reference that was updated	skills/bestax-theming/SKILL.md:27

Overall: The change is sound, and the diagnosis is right at the source: bulma-theme at node_modules/bulma/sass/utilities/css-variables.scss:489 is the only attribute hook Bulma has, and it is prefix-named; nothing in bulma-ui/src/scss adds a [data-theme] rule of its own. So a prefixed sheet genuinely carries no data-theme hook, and writing both attributes is both necessary and sufficient. The riskiest part is Theme.styles.test.tsx, the only thing standing between this and a silent regression — it leans on two non-obvious premises (jsdom applies an @media list only when it names screen, and Bulma emits its :root{light} default before the prefers-color-scheme blocks, so the renamed block wins the cascade), both of which I confirmed against bulma/sass/themes/_index.scss:11-34 and a green run. A human should look first at advisory 1: the reach of the fix is a React-context question, while the attribute it writes is a document-global one.

Residual risk:

• A prefixed sheet carrying a data-theme hook after all — refuted. bulma-theme (css-variables.scss:489-493) is the only attribute hook and it interpolates $class-prefix; a grep over bulma-ui/src/scss finds no [data-theme], theme-dark or prefers-color-scheme rule of our own. The extra data-theme written under a prefix is inert, not conflicting.
• The other Bulma hook, .<prefix>theme-dark — not a regression: colorMode never wrote a class, before or after, so the class path is equally unreachable on both branches.
• bestax-no-dark-mode.css — refuted as a gap. It emits :root { light-theme; setup-theme } and no [data-theme=…] block at all, so colorMode="dark" is a no-op there before and after. The filter in themedStylesheets() excludes it correctly (its SCSS matches neither import form the regex looks for), and an accidental inclusion would throw has no dark theme block in beforeAll rather than pass quietly.
• The invalid-prefix path throwing — refuted. getAttribute and removeAttribute do no name validation, so only setAttribute can throw and it is guarded; the setAttribute branch in the cleanup needs a non-null previous, which an invalid name can never have, because the HTML tokenizer terminates an attribute name at a slash. Exercised live by the md/ test (Theme.test.tsx:1118), unmount restore included.
• Nested Themes with colorMode — previous is still a per-effect snapshot, so two mounted Themes still restore last-in-wins. Unchanged in kind by this PR (one name became two, with identical per-name semantics), and the restore tests cover the single-Theme contract.
• Open: the ConfigProvider placement dependency of advisory 1. Mitigated, not closed, by convention — every shipped prefixed example puts ConfigProvider outermost (docs/docs/guides/getting-started/variations.md:67 and :152), so no documented arrangement hits it.

What I ran

Check	Result
jest src/helpers/__tests__/Theme*.test.tsx	106 passed
jest src/helpers (whole folder)	33 suites, 1038 passed
jest --coverage --collectCoverageFrom=src/helpers/Theme.tsx	100% stmts / branch / funcs / lines
tsc -p tsconfig.test.json --noEmit	clean
eslint on the three changed sources	clean
gen-mcp-index.mjs, gen-skills-rosters.mjs, gen-eslint-meta.mjs	all regenerate to an empty git status — no stale artefact

Reverting Theme.tsx in this checkout was not permitted, so I did not re-run the "fails on main" claim directly. It follows from the harness: for a bestax- prefix, the pre-fix effect writes only data-theme, for which the compiled prefixed sheet has no rule, so schemeMainL() resolves to the OS value and turns the page dark with colorMode="dark" on a light OS must fail.

🏄 Clean little set-wave, this one — paddled straight at the real reef (Bulma names its scheme attribute after the prefix, so the prefixed sheets were never listening) and rode it in with a styles test that compiles the actual SCSS instead of trusting an attribute merely landing in the DOM. Good to go, dude — just know the prefixed half of the fix only works when ConfigProvider sits upwind of your Theme.

[github-actions]

📸 Story screenshots at handoff — 81b02df

6 stories (light + dark) from this PR's changed files under bulma-ui/src/ (workflow run · artifact story-screenshots-pr-894, 30-day retention).

helpers-theme--sunset-theme · Helpers/Theme · Sunset Theme

Light	Dark

helpers-theme--forest-theme · Helpers/Theme · Forest Theme

Light	Dark

helpers-theme--global-theme · Helpers/Theme · Global Theme

Light	Dark

helpers-theme--local-theme · Helpers/Theme · Local Theme

Light	Dark

helpers-theme--theme-with-class-name · Helpers/Theme · Theme With Class Name

Light	Dark

helpers-theme--nested-themes · Helpers/Theme · Nested Themes

Light	Dark

> ✂ 4 more matched stories skipped by the cap: `helpers-theme--bulma-prefixed-theme`, `helpers-theme--shadow-theme`, `helpers-theme--radius-theme`, `helpers-theme--dark-mode` — full set in the artifact.

[allxsmith]

Every finding from the review, fixed or answered, in 75d504a:

1. The prefixed attribute depends on Theme sitting inside ConfigProvider: refuted. That's consistent with every other prefixed output. Theme reads the prefix from ConfigProvider context the way every component reads it for its classes, so a Theme mounted above the provider is outside the prefixed app just as an unprefixed Button there would be. The TSDoc and the Theme page say the prefixed attribute is written under a ConfigProvider classPrefix, and every documented prefixed setup puts the provider outermost.
2. A prefix that can't form an attribute name failed silently: fixed. It now raises a development warning naming the attribute and saying the prefixed sheet keeps its scheme; the test asserted the warning before the code had it.
3. The docs named only data-bestax-theme, or restated the mechanism: fixed in part. The theming skill's dark mode summary and the Theme page's props table name data-<prefix>theme, and the CSS variables guide links to the Theme page's dark mode section rather than restating it. The DarkMode story's text is left as is: it renders without a class prefix, so "sets data-theme on <html>" is exactly what it does there.

[allxsmith]

deep-review: fresh

The advisories from the last review are fixed or answered in the comment above; this reviews 75d504a.

[github-actions]

Preview Deployment

Preview URL: https://ee92fd41.bestax.pages.dev

[claude]

Deep review — 0 blocking · 4 advisory

#	Severity	Area	Finding	Location
1	🔵 Advisory	API	The new invalid-prefix development warning is described nowhere a consumer reads — not in the colorMode TSDoc, not in the Theme page's dark-mode prose or its props-table row — unlike Theme's other two dev warnings (isRoot radius, isRoot columnGap), which the props table calls out explicitly	bulma-ui/src/helpers/Theme.tsx:1101
2	🔵 Advisory	Correctness	Adjacent surfaces in the same failure class still name only the unprefixed attribute, and none is touched here: the PurgeCSS example safelists /data-theme/, which does not match data-bestax-theme (docs/docs/guides/getting-started/optimizing-css.md:78); the per-surface recipe is <Box data-theme="light"> (docs/docs/guides/helpers/color.md:158); plus docs/docs/skills/theming.mdx:192, skills/bestax-theming/examples/dark-mode.tsx:4 and bulma-ui/src/helpers/CLAUDE.md. Each is true of an unprefixed build — incomplete, not wrong	docs/docs/guides/getting-started/optimizing-css.md:78
3	🔵 Advisory	Robustness	The diagnostic is mode-dependent: removeAttribute never validates a name, so colorMode="system" under an attribute-invalid prefix warns nothing. Benign — there is no prefixed attribute to remove — but the misconfiguration only surfaces on 'light'/'dark'	bulma-ui/src/helpers/Theme.tsx:1090
4	🔵 Advisory	Coverage	The harness floor asserts only that '' and 'bestax-' appear among the discovered prefixes, so a sheet silently dropped by the content filter — the other unprefixed one, bestax-no-helpers.scss, say — would still pass it. All four themed sheets do run today (37 tests = 4 × 9 + 1)	bulma-ui/src/helpers/__tests__/Theme.styles.test.tsx:60

Overall: Sound, and the diagnosis holds up empirically rather than by argument — I compiled src/scss/versions/bestax-prefixed.scss and its only scheme hooks are [data-bestax-theme=light], [data-bestax-theme=dark], .bestax-theme-light and .bestax-theme-dark, with zero data-theme selectors anywhere in the sheet, so a prefixed build genuinely was deaf to colorMode and writing both names is both necessary and sufficient. The riskiest part remains Theme.styles.test.tsx — it is the only thing standing between this and a silent regression, and it rests on two unasserted premises (jsdom honours an @media list only when it names screen; Bulma emits its prefers-color-scheme blocks before the attribute blocks, so the attribute wins at equal specificity) — both of which I confirmed against the compiled cascade. A human can merge on the code and spend their attention on advisory 2 instead: the executable fix is complete, but the surrounding guidance a reader or an agent follows for a prefixed app still points at data-theme alone, and no check in this repo can see that.

Residual risk:

• A prefixed sheet carrying a data-theme hook after all — refuted empirically. Compiled bestax-prefixed.scss with the real load path; a grep for the [data-…theme=…] form returns only the two data-bestax-theme selectors, and the class hooks are .bestax-theme-*. So the extra data-theme written under a prefix is inert, never conflicting.
• The harness quietly skipping a sheet — refuted. The content filter admits bestax.scss (''), bestax-no-helpers.scss (''), bestax-prefixed.scss (bestax-) and bestax-no-helpers-prefixed.scss (bestax-), and correctly rejects extras.scss (no Bulma import) and bestax-no-dark-mode.scss (imports only bulma/sass/utilities and bulma/sass/themes/light, neither of which the regex matches). --verbose confirms 37 tests, i.e. all four sheets × 9 cases plus the floor. An accidental inclusion of the no-dark-mode sheet would throw has no dark theme block in beforeAll, not pass quietly.
• The cleanup's unguarded root.setAttribute(name, value) — refuted. previous[i] for an attribute-invalid name is always null: setAttribute is the only route that creates one and it is the guarded call, and HTML's attribute-name tokenizer terminates a name at /, so markup cannot produce data-md/theme either. The restore therefore takes the removeAttribute branch, which performs no name validation. Exercised live by the md/ test's post-unmount assertion.
• A case-differing prefix (classPrefix="Bestax-") — refuted. setAttribute ASCII-lowercases the qualified name on an HTML element, and attribute names in selectors are matched ASCII case-insensitively in HTML documents, so Bulma's [data-Bestax-theme=dark] still matches the data-bestax-theme that lands. No silent half-failure between the two casings.
• classPrefix changing while a colorMode Theme is mounted — refuted. classPrefix is in the effect's dep list, so the old name set is restored before the new one is written; useClassPrefix returns a plain string from context, so the effect does not re-run spuriously.
• Test pollution from the restore-on-unmount contract — refuted, and the new explicit cleanup() is load-bearing rather than decorative. RTL registers its auto-cleanup afterEach at root level when the module is imported, and Jest runs afterEach innermost-first, so without the explicit call the data-bestax-theme=light pre-set test would have its value re-added by the effect cleanup after the hook removed it.
• Nested Themes both setting colorMode — unchanged in kind. previous is still a per-effect snapshot, so two mounted Themes still restore last-in-wins; one name became two with identical per-name semantics.
• Open, in an executable path: only advisory 3's asymmetry — 'system' never warns. Everything else in the failure class that this diff touches is closed above; the remaining exposure is the unprefixed-only guidance of advisory 2, which is prose and so out of scope here.

What I ran

Check	Result
jest Theme.test.tsx Theme.styles.test.tsx	2 suites, 106 passed
jest src/helpers/__tests__/Theme (incl. SSR)	3 suites, 108 passed
jest src/helpers (whole folder)	33 suites, 1038 passed
jest --listTests src/helpers	Theme.styles.test.tsx is in the default run, not orphaned
jest Theme.styles.test.tsx --verbose	37 tests — all four themed sheets exercised
jest --coverage --collectCoverageFrom=src/helpers/Theme.tsx	100% stmts / branch / funcs / lines
pnpm typecheck:tests (tsc -p tsconfig.test.json)	clean
eslint on the three changed sources	clean
prettier --check on all seven hand-edited files	clean
pnpm check:conformance	all 23 checks pass
gen:mcp:check, gen:catalog:check	both regenerate identically — no stale artefact
sass compile of bestax-prefixed.scss, selector grep	only [data-bestax-theme=…] / .bestax-theme-*; no data-theme hook
Cascade order in that compiled sheet	prefers-color-scheme at 916/1784, [data-bestax-theme=…] at 1834/2735 — the attribute wins at equal specificity, so the harness's rename trick is valid
Build preview site check on this SHA	success, and onBrokenAnchors: 'throw' — so the new theme.md#dark-mode link resolves

I could not write to the checkout's sources, so the "fails on main" claim was not re-run by reverting the effect. The compiled-sheet evidence makes it a certainty for the two prefixed sheets: the pre-fix effect writes only data-theme, for which those sheets contain no rule at all.

Housekeeping, not a PR change: my sass invocation left an untracked bulma-ui/bulma-ui/.tmp-pref.css that the sandbox would not let me delete. It is mine, not this PR's — git clean -fdx bulma-ui/bulma-ui removes it.

🏄 Second lap on this one and it's cleaner than the first — the silent catch now actually hollers when a gnarly prefix eats the attribute, and the docs stopped pretending data-theme is the only name in the water. Good to go, dude; the only kelp left is the guidance around the fix, like that PurgeCSS safelist still only knowing the unprefixed name.

[allxsmith]

deep-review: fresh

Main is merged in, so this review posts every finding, advisory included, as its own thread. It reviews 0e084e0.

[github-actions]

Preview Deployment

Preview URL: https://3179c778.bestax.pages.dev

[claude]

Deep review — 0 blocking · 4 advisory

#	Severity	Area	Finding	Location
1	🔵 Advisory	Correctness	Four adjacent colorMode surfaces still name only data-theme; the PurgeCSS safelist is the one this PR makes worse — the prefixed name is now assembled at runtime, so data-bestax-theme is a literal in no scanned file and /data-theme/ cannot match it	skills/bestax-theming/SKILL.md:27
2	🔵 Advisory	API	The new invalid-prefix dev warning is documented nowhere a consumer reads, unlike Theme's other three dev warnings, which the props table calls out — and helpers/theme.md is in GENERATED_EXEMPT, so nothing will reconcile the two surfaces	bulma-ui/src/helpers/Theme.tsx:856
3	🔵 Advisory	Robustness	The diagnostic is mode-dependent: removeAttribute never validates a name, so colorMode="system" — the initial state the shipped skill example uses — warns nothing under an attribute-invalid prefix	bulma-ui/src/helpers/Theme.tsx:1123
4	🔵 Advisory	Coverage	The harness floor asserts prefix membership, so a themed sheet dropped by the content regex still passes while a sibling supplies its prefix	bulma-ui/src/helpers/__tests__/Theme.styles.test.tsx:62

Overall: The fix is correct and the diagnosis is right at the source. @mixin bulma-theme (node_modules/bulma/sass/utilities/css-variables.scss:489-493) emits [data-#{$class-prefix}theme="…"], .#{$class-prefix}theme-… and is Bulma's only attribute hook; a grep for data-theme across all of bulma/sass and bulma-ui/src/scss returns two comment lines and nothing else, so the prefixed sheets genuinely carried no data-theme rule and writing both names is both necessary and sufficient. The riskiest artefact is Theme.styles.test.tsx — it is the only thing standing between this and a silent regression, and it leans on two unasserted premises (jsdom honours an @media list only when it names screen; Bulma emits its prefers-color-scheme blocks before the attribute blocks, so the renamed block loses to the attribute at equal specificity), both of which the green run confirms. A human can merge on the code and spend their attention on finding 1 instead: the executable fix is complete, but the guidance around it that a prefixed app follows still points at the unprefixed name, and no check here can see that.

Residual risk:

• A prefixed sheet carrying a data-theme hook after all — refuted at the source. bulma-theme interpolates $class-prefix and is the only attribute hook; grep -rn data-theme node_modules/bulma/sass bulma-ui/src/scss yields only themes/_index.scss:26 and :32, both comments. The extra data-theme written under a prefix is inert, never conflicting — and it is not new behaviour, so no prefixed app gains a collision it did not already have.
• The other hook, .<prefix>theme-dark — not a regression. colorMode wrote no class before this PR and writes none after, so the class path is equally unreachable on both branches.
• bestax-no-dark-mode.css — refuted as a gap. It loads only bulma/sass/themes/light, so it has no [data-…theme=dark] block at all and colorMode="dark" was a no-op there before and after. The filter excludes it correctly, and an accidental inclusion throws has no dark theme block in beforeAll rather than passing quietly.
• The invalid-prefix path throwing anywhere — refuted. Only setAttribute validates a name, and it is the guarded call; getAttribute returns null and removeAttribute is a no-op for a name no element can hold, so the cleanup always takes the removeAttribute branch (previous[i] cannot be non-null for an attribute-invalid name). Exercised live by the md/ test, post-unmount assertion included.
• classPrefix changing under a mounted colorMode Theme — refuted. It is in the effect's dep list, so the old name set is restored before the new one is written, and useClassPrefix returns a plain string from context, so the effect does not churn.
• A case-differing prefix (classPrefix="Bestax-") — refuted. setAttribute ASCII-lowercases the qualified name on an HTML element, and selector attribute names match ASCII case-insensitively in HTML documents, so Bulma's [data-Bestax-theme=dark] still matches the data-bestax-theme that lands.
• Nested Themes both setting colorMode — unchanged in kind. previous is still a per-effect snapshot, so two mounted Themes restore last-in-wins. One name became two with identical per-name semantics; the restore tests cover the single-Theme contract.
• A Theme colorMode mounted above its ConfigProvider — settled, not re-raised. Answered in the author's comment on the first review as deliberate and consistent with every other prefixed output, with the TSDoc and Theme page both scoping the prefixed attribute to "under a ConfigProvider classPrefix"; the code it points at has not changed, so the reason still holds.
• Open: finding 3's 'system' asymmetry is the only executable path in the failure class still unclosed. Everything else above is refuted; finding 1 is prose and so not a residual risk by definition, but it is where the remaining real-world exposure sits.

What I ran

Check	Result
jest Theme.test.tsx Theme.styles.test.tsx	2 suites, 111 passed
jest Theme.styles.test.tsx --verbose	37 tests — all four themed sheets x 9 cases, plus the floor
jest --coverage --collectCoverageFrom=src/helpers/Theme.tsx	100% stmts / branch / funcs / lines
pnpm check:conformance	all 23 checks pass
pnpm gen:mcp:check	regenerates identically — no stale MCP index
pnpm gen:catalog:check	regenerates identically — no stale catalog
prettier --check on all seven hand-edited files	clean
grep -rn data-theme node_modules/bulma/sass bulma-ui/src/scss	two comment lines only — no unprefixed attribute hook anywhere
bulma/sass/utilities/css-variables.scss:489-493 read directly	[data-#{$class-prefix}theme] / .#{$class-prefix}theme-… confirmed as the sole hooks
src/scss flavour-by-flavour trace against the test's content regex	4 admitted, 2 correctly rejected (table in finding 4)
scripts/lib/api-sources.mjs	helpers/ is in GENERATED_EXEMPT — theme.md's props table is hand-written, so finding 2's two surfaces drift independently
Checks on this SHA	Build and Test, Test and Preview Deployment, Build preview site, React 18 + 19, CodeQL all success

Reverting the effect to re-run the "fails on main" claim was not permitted in this checkout (both file writes and git show redirection were blocked), so it rests on the source evidence above rather than a red run: for a bestax- prefix the pre-fix effect wrote only data-theme, for which those two compiled sheets contain no rule at all, so schemeMainL() necessarily resolved to the OS value.

🏄 Third lap and the wave's still holding its shape — the reef was real (Bulma names its scheme attribute after the prefix, so the prefixed sheets were never listening) and the styles test paddles out and compiles the actual SCSS instead of trusting an attribute merely landing in the DOM. Good to go, dude; the only kelp left is the guidance around the fix, like that PurgeCSS safelist still only knowing the unprefixed name.

[allxsmith]

deep-review: verify

Every open thread has a reply: fixed, with the commit, or refuted with a reason. This settles them on 81b02df.

[github-actions]

Preview Deployment

Preview URL: https://8ad8f651.bestax.pages.dev

[claude]

Deep review (verify) — 0 resolved · 0 open

No threads left to settle — all four threads from this review were already resolved on this same commit (81b02dfe), each carrying its verification or acceptance reply:

Thread	Severity	Disposition
bulma-ui/src/helpers/Theme.tsx:859 — new dev warning documented nowhere	🔵 Advisory	Verified fixed (ece8511a), resolved
bulma-ui/src/helpers/Theme.tsx:1129 — the system mode never warns	🔵 Advisory	Refutation accepted, resolved
bulma-ui/src/helpers/__tests__/Theme.styles.test.tsx — harness floor	🔵 Advisory	Verified fixed (ece8511a), resolved
skills/bestax-theming/SKILL.md:30 — adjacent colorMode surfaces	🔵 Advisory	Verified fixed (ece8511a, 81b02dfe), resolved

Overall: This pass settled threads and reviewed no commits. Re-reading the current tree confirms every disposition still holds: the colorMode TSDoc (Theme.tsx:860-861) and the props row (docs/docs/api/helpers/theme.md:481) both scope the warning to light/dark, the stylesheet floor asserts the exact complement of the content filter (Theme.styles.test.tsx:64-72), and the PurgeCSS safelist pattern in both deep and greedy now matches the prefixed attribute name as well as the plain one (optimizing-css.md:78-79). Theme.styles.test.tsx + Theme.test.tsx: 111 passed. Nothing of mine is open, so a human can go straight to the merge decision; anyone wanting the current code looked at again should ask for a fresh review rather than another verify.

Residual risk: none identified by this pass — a verify pass raises no findings, and the residual risks for this change were enumerated in the fresh review of 0e084e0b and answered in the four threads above.

🏄 Nothing left in the lineup on this one, dude — every wave already got ridden clean, threads closed, tests green. Just a mellow paddle back to shore.

[bestax-release-bot]

🎉 This PR is included in version 5.26.3 🎉

The release is available on:

• npm package (@latest dist-tag)
• GitHub release

Your semantic-release bot 📦🚀

[bestax-release-bot]

🎉 This PR is included in version 2.25.0 🎉

The release is available on:

• npm package (@latest dist-tag)
• GitHub release

Your semantic-release bot 📦🚀

[bestax-release-bot]

🎉 This PR is included in version 4.2.13 🎉

The release is available on:

• npm package (@latest dist-tag)
• GitHub release

Your semantic-release bot 📦🚀

[bestax-release-bot]

🎉 This PR is included in version 1.14.1 🎉

The release is available on:

• npm package (@latest dist-tag)
• GitHub release

Your semantic-release bot 📦🚀