@erclx/canon 5.62.0 → 5.63.0

package/governance/ledger/{ui → skills/ui-craft}/motion.md

@@ -1,9 +1,9 @@
 ---
-title: Motion rule ledger
-description: Sources for governance/rules/ui/motion.md, one row per bullet
+title: Motion reference ledger
+description: Sources for claude/skills/ui-craft/references/motion.md, one row per bullet
 ---
 
-# Motion rule ledger
+# Motion reference ledger
 
 The field is web accessibility, and the authorities are the W3C, through WCAG 2.2 and the ARIA Authoring Practices Guide, and MDN for the media feature.
 

package/governance/ledger/{ui → skills/ui-craft}/pointer.md

@@ -1,9 +1,9 @@
 ---
-title: Pointer rule ledger
-description: Sources for governance/rules/ui/pointer.md, one row per bullet
+title: Pointer reference ledger
+description: Sources for claude/skills/ui-craft/references/pointer.md, one row per bullet
 ---
 
-# Pointer rule ledger
+# Pointer reference ledger
 
 The field is web accessibility, and the authorities are the W3C, through WCAG 2.2 and the ARIA Authoring Practices Guide, MDN for popover behavior, and the platform guidelines from Apple and Android for native targets. These bullets were split from `keyboard.md`.
 

package/governance/ledger/{ui/ux-completeness.md → skills/ui-craft/states.md}

@@ -1,9 +1,9 @@
 ---
-title: UX completeness rule ledger
-description: Sources for governance/rules/ui/ux-completeness.md, one row per bullet
+title: UX completeness reference ledger
+description: Sources for claude/skills/ui-craft/references/states.md, one row per bullet
 ---
 
-# UX completeness rule ledger
+# UX completeness reference ledger
 
 The field is interaction design, and the authorities are the Nielsen Norman Group's research articles, WCAG 2.2 for timing, MDN for the `title` attribute and CSS truncation, and web.dev for layout shift.
 

package/governance/ledger/ui/craft.md

@@ -0,0 +1,19 @@
+---
+title: UI craft rule ledger
+description: Sources for governance/rules/ui/craft.md, one row per bullet
+---
+
+# UI craft rule ledger
+
+The rule is a pointer to the `ui-craft` and `design-taste` skills, which carry their own sources, so both rows are house choices naming the bug the pointer prevents.
+
+## Authority
+
+| Bullet                     | Source                                                                                                                                                                                      | Read       |
+| -------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------- |
+| Load the skills            | House choice: a rendered edit made without the behavior floor or the layer order reaches for a model's defaults, and a second copy of either in the rule drifts from the skill that owns it | 2026-10-24 |
+| Report an unresolved skill | House choice: the rule ships with the CLI and the skills with the plugin, so a project holding governance alone would otherwise proceed silently without them                               | 2026-10-18 |
+
+## Declined
+
+The fold kept both bullets and widened the first to name `ui-craft` beside `design-taste`, since the practice rules it replaced now live in that skill.

package/governance/renames.toml

@@ -3,9 +3,11 @@
 # so `canon gov sync` replaces a held old file with the rule under its new name
 # rather than only deleting it.
 #
-# An entry declares one-to-one identity and nothing else. A fold of several
-# rules into one, or a split of one into several, is a retirement plus an
-# addition and takes no entry here. Add the entry in the same change as the
+# An entry declares one-to-one identity, or a fold onto a pointer rule that
+# now routes to the skill carrying the folded rules, so a target holding any
+# of them lands on the pointer rather than losing the practice with nothing in
+# its place. Any other fold of several rules into one, or a split of one into
+# several, is a retirement plus an addition and takes no entry here. Add the entry in the same change as the
 # rename or renumber. Prune an entry only once its target rule is retired and
 # the retire path covers it, since a chain ending on a rule nothing ships
 # resolves to nothing and a sync deletes the held file instead. Otherwise leave
@@ -73,3 +75,15 @@
 "460-design-taste" = "design-taste"
 "465-interface-casing" = "interface-casing"
 "470-motion" = "motion"
+"design-taste" = "craft"
+"a11y" = "craft"
+"keyboard" = "craft"
+"pointer" = "craft"
+"forms" = "craft"
+"motion" = "craft"
+"ux-controls" = "craft"
+"ux-completeness" = "craft"
+"surface-capture" = "craft"
+"copy" = "craft"
+"interface-casing" = "craft"
+"link-behavior" = "craft"

package/governance/rules/lib/test-scope.md

@@ -20,5 +20,5 @@ paths:
 
 - Answer a question about the running page with a script against the dev server rather than with the suite.
 - Do not enable `fullyParallel` in the Playwright config.
-- Follow `.claude/rules/canon/ui/surface-capture.md` for capture scope.
+- Follow the `canon:ui-craft` skill's capture reference for capture scope, reporting it when the skill does not resolve.
 - Follow `.claude/rules/canon/lib/e2e-reliability.md` for waits and guards.

package/governance/rules/ui/craft.md

@@ -0,0 +1,22 @@
+---
+description: Route edits to a rendered surface to the ui-craft skill for how it behaves and to the design-taste skill for how it looks
+paths:
+  - '**/*.tsx'
+  - '**/*.jsx'
+  - '**/*.vue'
+  - '**/*.svelte'
+  - '**/*.astro'
+  - '**/*.html'
+  - '**/*.mdx'
+  - '**/*.css'
+  - '**/*.scss'
+  - '**/content/**/*.ts'
+  - 'canon/DESIGN.md'
+---
+
+# UI craft standards
+
+## Authority
+
+- Load the `canon:ui-craft` skill for how a surface behaves and the `canon:design-taste` skill for how it looks. Each is the single source for its half.
+- Report it rather than proceeding silently when either skill does not resolve. Both ship with the plugin and this rule ships with the CLI, so governance alone carries neither.

package/governance/splits.toml

@@ -11,5 +11,3 @@
 # entry only once its key or partner is retired, which the live-rule check in
 # `src/gov/splits.test.ts` forces.
 [split]
-keyboard = ["pointer"]
-ux-completeness = ["ux-controls"]

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@erclx/canon",
   "type": "module",
-  "version": "5.62.0",
+  "version": "5.63.0",
   "description": "Runs your agent's work and keeps its code good, on one setup for every repo.",
   "license": "MIT",
   "bin": {

package/src/claude/cases/misc.ts

@@ -17,6 +17,11 @@ export const MISC_CASES: readonly SkillCase[] = [
       'Which layer of this landing page should I settle first, and what has to stay constant across it?',
     expect: 'design-taste',
   },
+  {
+    prompt:
+      'This settings form needs validation and an empty state, so what does it owe for keyboard users and error messages before I build it?',
+    expect: 'ui-craft',
+  },
   {
     prompt:
       'This project has no logo yet. Draft one and give me a social card to go with it.',

package/src/claude/skills-audit.ts

@@ -1,8 +1,20 @@
 import { existsSync } from 'node:fs'
-import { dirname, join } from 'node:path'
+import { dirname, join, relative } from 'node:path'
 import { isFamilyKey } from '@/claude/skills-families'
+import { plural } from '@/cli/ui'
+import {
+  compareLedger,
+  countLedgerRows,
+  type LedgerMismatch,
+  measureRule,
+} from '@/gov/audit'
 import { parseFrontmatter, readField } from '@/indexes/frontmatter'
-import { type BodyLine, bodyLines, maskDisplayed } from '@/markdown/scan'
+import {
+  type BodyLine,
+  bodyLines,
+  maskCodeSpans,
+  maskDisplayed,
+} from '@/markdown/scan'
 
 /** Returned when a skill folder carries no `REQUIREMENT.md`, the gating check. */
 export const EXIT_MISSING_REQUIREMENT = 2
@@ -69,8 +81,22 @@ export const LEDGER_SKILLS: readonly string[] = [
   join(SHIPPED_CORPUS, 'design-taste'),
   join(SHIPPED_CORPUS, 'write-human'),
   join(SHIPPED_CORPUS, 'test-first'),
+  join(SHIPPED_CORPUS, 'ui-craft'),
 ]
 
+/**
+ * The ledger-listed skills whose every reference but the source record carries
+ * a ledger of its own, one table row per bullet under each H2, the way a rule
+ * in a practice folder does. The ledgers sit outside the plugin as rule
+ * ledgers do, so a target pays nothing for them.
+ */
+export const BULLET_LEDGER_SKILLS: readonly string[] = [
+  join(SHIPPED_CORPUS, 'ui-craft'),
+]
+
+/** Where a per-bullet ledger lives, as `<skill>/<reference>.md` below it. */
+export const SKILL_LEDGERS = join('governance', 'ledger', 'skills')
+
 /** The closing H2s a practice skill carries, matched exactly outside fences. */
 export const PRACTICE_SECTIONS: readonly string[] = [
   'Excuses and rebuttals',
@@ -81,6 +107,19 @@ export const PRACTICE_SECTIONS: readonly string[] = [
 /** The source ledger every practice skill carries beside its body. */
 export const PRACTICE_LEDGER = join('references', 'adopted.md')
 
+/**
+ * A ledger cites a source by linking a page outside the project, as an inline
+ * link or an autolink. A relative link points back into the skill and cites
+ * nothing a reader could check the rule against.
+ */
+const EXTERNAL_LINK = /\]\(\s*https?:\/\/|<https?:\/\//
+
+/**
+ * A table cell opening on the marker the governance ledgers already use for a
+ * rule the toolkit chose, so one check reads every ledger alike.
+ */
+const HOUSE_CHOICE = /\|\s*House choice\b/
+
 /**
  * The line count past which a reference opens with a contents list, from
  * Anthropic's skill authoring guidance. Counted over the whole file the way
@@ -166,13 +205,23 @@ interface SkillSource {
   readonly requirement: string | undefined
   /** Every H2 in `SKILL.md` outside a fence, for the practice shape. */
   readonly sections: readonly string[]
-  readonly hasLedger: boolean
+  /** Undefined when the folder carries no ledger at all. */
+  readonly ledger: string | undefined
   /**
    * `SKILL.md` and every reference below the folder, keyed by the path a
    * finding names. `REQUIREMENT.md` and `EVAL.md` are left out, since the
    * requirement is where the standard sends a date.
    */
   readonly bodies: readonly SkillText[]
+  /** One entry per reference of a skill on the per-bullet ledger list. */
+  readonly referenceLedgers: readonly ReferenceLedger[]
+}
+
+interface ReferenceLedger {
+  readonly rel: string
+  readonly ledgerRel: string
+  /** Undefined when the ledger file is absent. */
+  readonly mismatches: readonly LedgerMismatch[] | undefined
 }
 
 interface SkillText {
@@ -195,7 +244,7 @@ export async function auditSkills(root: string): Promise<SkillsAudit> {
   )
 
   const perCorpus = await Promise.all(
-    present.map((corpus) => readCorpus(corpus.rel, corpus.dir)),
+    present.map((corpus) => readCorpus(root, corpus.rel, corpus.dir)),
   )
   const sources = perCorpus.flat()
 
@@ -247,21 +296,29 @@ export function auditExitCode(report: SkillsAudit): number {
   return report.missingRequirement.length > 0 ? EXIT_MISSING_REQUIREMENT : 0
 }
 
-async function readCorpus(rel: string, dir: string): Promise<SkillSource[]> {
+async function readCorpus(
+  root: string,
+  rel: string,
+  dir: string,
+): Promise<SkillSource[]> {
   const paths = [
     ...new Bun.Glob('*/SKILL.md').scanSync({ cwd: dir, onlyFiles: true }),
   ].sort()
 
-  return Promise.all(paths.map((path) => readSkill(rel, dir, dirname(path))))
+  return Promise.all(
+    paths.map((path) => readSkill(root, rel, dir, dirname(path))),
+  )
 }
 
 async function readSkill(
+  root: string,
   rel: string,
   dir: string,
   folder: string,
 ): Promise<SkillSource> {
   const skillDir = join(dir, folder)
   const requirementPath = join(skillDir, 'REQUIREMENT.md')
+  const ledgerPath = join(skillDir, PRACTICE_LEDGER)
 
   const references = [
     ...new Bun.Glob('references/**/*.md').scanSync({
@@ -270,11 +327,14 @@ async function readSkill(
     }),
   ].sort()
 
-  const [body, requirement, referenceTexts] = await Promise.all([
+  const [body, requirement, ledger, referenceTexts] = await Promise.all([
     Bun.file(join(skillDir, 'SKILL.md')).text(),
     existsSync(requirementPath)
       ? Bun.file(requirementPath).text()
       : Promise.resolve(undefined),
+    existsSync(ledgerPath)
+      ? Bun.file(ledgerPath).text()
+      : Promise.resolve(undefined),
     Promise.all(
       references.map((path) => Bun.file(join(skillDir, path)).text()),
     ),
@@ -282,6 +342,15 @@ async function readSkill(
 
   const fields = parseFrontmatter(body)
   const skillRel = join(rel, folder)
+  const referenceLedgers = BULLET_LEDGER_SKILLS.includes(skillRel)
+    ? await readReferenceLedgers(
+        root,
+        skillRel,
+        folder,
+        references,
+        referenceTexts,
+      )
+    : []
 
   return {
     rel: skillRel,
@@ -294,7 +363,7 @@ async function readSkill(
       requirement === undefined ? undefined : headings(requirement),
     requirement,
     sections: sectionsOf(body),
-    hasLedger: existsSync(join(skillDir, PRACTICE_LEDGER)),
+    ledger,
     bodies: [
       { rel: join(skillRel, 'SKILL.md'), text: body, isReference: false },
       ...references.map((path, index) => ({
@@ -303,9 +372,45 @@ async function readSkill(
         isReference: true,
       })),
     ],
+    referenceLedgers,
   }
 }
 
+/**
+ * Pairs each reference's bullets with its ledger's rows through the rule
+ * audit's own comparison, so a reference and a rule are held to one count.
+ * The source record names authorities rather than stating bullets, so it is
+ * the one reference asked for no ledger.
+ */
+async function readReferenceLedgers(
+  root: string,
+  skillRel: string,
+  folder: string,
+  references: readonly string[],
+  texts: readonly string[],
+): Promise<ReferenceLedger[]> {
+  return Promise.all(
+    references
+      .map((path, index) => ({ path, text: texts[index] ?? '' }))
+      .filter(({ path }) => path !== PRACTICE_LEDGER)
+      .map(async ({ path, text }) => {
+        const ledgerRel = join(
+          SKILL_LEDGERS,
+          folder,
+          relative('references', path),
+        )
+        const ledgerPath = join(root, ledgerRel)
+        const mismatches = existsSync(ledgerPath)
+          ? compareLedger(
+              measureRule(text).groups,
+              countLedgerRows(await Bun.file(ledgerPath).text()),
+            )
+          : undefined
+        return { rel: join(skillRel, path), ledgerRel, mismatches }
+      }),
+  )
+}
+
 /**
  * A key carrying an empty value declares nothing, so blank reads as absent
  * rather than as a name that disagrees with every folder.
@@ -493,11 +598,44 @@ function practiceFindings(source: SkillSource): SkillFinding[] {
       rel: source.rel,
       detail: `missing section: ${section}`,
     }))
-  const ledger = source.hasLedger
-    ? []
-    : [{ rel: source.rel, detail: `missing ledger: ${PRACTICE_LEDGER}` }]
 
-  return [...sections, ...ledger]
+  return [
+    ...sections,
+    ...ledgerFindings(source),
+    ...source.referenceLedgers.flatMap(referenceLedgerFindings),
+  ]
+}
+
+function referenceLedgerFindings(entry: ReferenceLedger): SkillFinding[] {
+  if (entry.mismatches === undefined) {
+    return [{ rel: entry.rel, detail: `missing ledger: ${entry.ledgerRel}` }]
+  }
+  return entry.mismatches.map((mismatch) => ({
+    rel: entry.rel,
+    detail: `${mismatch.heading === null ? 'above the first H2' : `## ${mismatch.heading}`}: ${plural(mismatch.bullets, 'bullet')}, ${plural(mismatch.rows, 'ledger row')}`,
+  }))
+}
+
+/**
+ * A ledger present but citing nothing passed for as long as the check read the
+ * file's existence alone, so the contents are read too. A sample block shows a
+ * reader the shape of a row and cites nothing, so fences and code spans are
+ * left out.
+ */
+function ledgerFindings(source: SkillSource): SkillFinding[] {
+  if (source.ledger === undefined) {
+    return [{ rel: source.rel, detail: `missing ledger: ${PRACTICE_LEDGER}` }]
+  }
+
+  const cites = bodyLines(source.ledger)
+    .filter((line) => !line.fenced)
+    .map((line) => maskCodeSpans(line.text))
+    .some((text) => EXTERNAL_LINK.test(text) || HOUSE_CHOICE.test(text))
+  if (cites) return []
+
+  return [
+    { rel: source.rel, detail: `ledger cites nothing: ${PRACTICE_LEDGER}` },
+  ]
 }
 
 /**

package/src/commands/claude.ts

@@ -30,6 +30,8 @@ import {
   DESCRIPTION_LIMIT,
   LEDGER_SKILLS,
   PRACTICE_LEDGER,
+  BULLET_LEDGER_SKILLS,
+  SKILL_LEDGERS,
   PRACTICE_SECTIONS,
   PRACTICE_SKILLS,
   REQUIREMENT_LIMITS,
@@ -1768,6 +1770,9 @@ function reportPracticeShape(report: SkillsAudit): void {
   logInfo(
     `A listed practice skill closes with ${PRACTICE_SECTIONS.map((section) => `## ${section}`).join(', ')}, and every skill on the wider ledger list carries ${PRACTICE_LEDGER}.`,
   )
+  logInfo(
+    `A skill on the per-bullet list, ${plural(BULLET_LEDGER_SKILLS.length, 'skill')} here, also matches each other reference to a ledger under ${SKILL_LEDGERS}, one row per bullet under each H2.`,
+  )
   logInfo(
     `Both lists are kept here and name ${plural(PRACTICE_SKILLS.length, 'practice skill')} and ${plural(LEDGER_SKILLS.length, 'ledger skill')}, so a skill outside ${SHIPPED_CORPUS} is never read.`,
   )

package/src/commands/tooling.ts

@@ -258,7 +258,7 @@ export function register(program: Command): void {
         'never overwrites one.',
         '',
         'Examples:',
-        '  canon tooling customize e2e/home.spec.ts',
+        '  canon tooling customize e2e/screenshot.ts',
         '',
       ].join('\n'),
     )
@@ -1014,6 +1014,9 @@ function reportPackage(result: ScanResult): void {
     logStep('Scanning dependencies')
     for (const entry of result.deps) {
       if (entry.state === 'present') logInfo(entry.name)
+      if (entry.state === 'ahead') {
+        logInfo(`${entry.name} ${entry.installed} (ahead of ${entry.spec})`)
+      }
     }
     for (const entry of result.deps) {
       if (entry.state === 'missing') logAdd(entry.spec)

package/src/gate/measures.ts

@@ -12,6 +12,7 @@ import {
   auditSkills,
   LEDGER_SKILLS,
   PRACTICE_LEDGER,
+  SKILL_LEDGERS,
   PRACTICE_SKILLS,
   SHIPPED_CORPUS,
 } from '@/claude/skills-audit'
@@ -384,7 +385,7 @@ export const skillProvenance: Measure = async (ctx) => {
 
 /**
  * Every listed practice skill against the closing sections, and every skill on
- * the wider ledger list against the ledger, both of which
+ * the wider ledger list against a ledger that cites something, both of which
  * `claude/skills/create-skill/references/skill-practice.md` requires.
  *
  * Read in-process for the reason `skillProvenance` is, so the verb keeps its
@@ -407,14 +408,14 @@ export const skillPracticeShape: Measure = async (ctx) => {
       .join(', ')
     return {
       emissions: [],
-      failure: `${plural(found.length, 'practice shape finding')}: ${where}. Add the closing H2 exactly as named to a practice skill, or the ledger at ${PRACTICE_LEDGER} to any skill on the ledger list.`,
+      failure: `${plural(found.length, 'practice shape finding')}: ${where}. Add the closing H2 exactly as named to a practice skill, or the ledger at ${PRACTICE_LEDGER} to any skill on the ledger list, citing one linked source or one House choice row. A reference on the per-bullet list takes one ledger row per bullet under the same H2 at ${SKILL_LEDGERS}.`,
     }
   }
 
   return {
     emissions: [
       info(
-        `${plural(PRACTICE_SKILLS.length, 'practice skill')} carry every closing section, and ${plural(LEDGER_SKILLS.length, 'listed skill')} carry a ledger`,
+        `${plural(PRACTICE_SKILLS.length, 'practice skill')} carry every closing section, and ${plural(LEDGER_SKILLS.length, 'listed skill')} carry a ledger citing a source`,
       ),
     ],
   }

package/src/gov/audit.ts

@@ -294,7 +294,12 @@ export function countLedgerRows(text: string): HeadingCount[] {
   return [...counts].map(([key, count]) => ({ heading: key, count }))
 }
 
-function compareLedger(
+/**
+ * Pairs bullets and ledger rows under each heading, naming every heading whose
+ * counts differ. The skill audit reads a practice skill's references through
+ * it, so a rule and a reference are held to one comparison.
+ */
+export function compareLedger(
   groups: readonly HeadingCount[],
   rows: readonly HeadingCount[],
 ): LedgerMismatch[] {

package/src/inventory/subjects.ts

@@ -27,7 +27,7 @@ export interface Subject {
 
 /**
  * Reports what each element changes about itself when it takes focus, which is
- * the reading `governance/rules/ui/keyboard.md` states three rules against and
+ * the reading the `ui-craft` keyboard reference states three rules against and
  * nothing measures.
  *
  * The difference between rest and focus is the treatment, rather than the

package/src/tooling/package.ts

@@ -17,7 +17,9 @@ export interface ScriptState {
 export interface DepState {
   readonly spec: string
   readonly name: string
-  readonly state: 'missing' | 'present'
+  readonly state: 'missing' | 'present' | 'ahead'
+  /** The spec the target records, set only when it is `ahead` of the pin. */
+  readonly installed?: string
 }
 
 export function readPackage(path: string): PackageJson | undefined {
@@ -89,18 +91,24 @@ function firstInteger(value: string): string | undefined {
 
 /**
  * Compares a manifest pin against what package.json records. A spec counts as
- * missing when absent, or when the pinned major differs from the installed
- * major. Anything looser than a major mismatch is left alone.
+ * missing when absent or on an earlier major than the pin, and ahead when it
+ * sits on a later one, which a write leaves alone. A spec with no integer to
+ * compare, such as `latest` or `workspace:*`, counts as present.
  */
-function isSatisfied(installed: string | undefined, pin: string): boolean {
-  if (installed === undefined) return false
-  if (pin === '') return true
+function compareSpec(
+  installed: string | undefined,
+  pin: string,
+): DepState['state'] {
+  if (installed === undefined) return 'missing'
+  if (pin === '') return 'present'
 
   const pinMajor = firstInteger(pin)
   const installedMajor = firstInteger(installed)
-  if (pinMajor === undefined) return true
+  if (pinMajor === undefined || installedMajor === undefined) return 'present'
 
-  return pinMajor === installedMajor
+  const delta = Number(installedMajor) - Number(pinMajor)
+  if (delta === 0) return 'present'
+  return delta > 0 ? 'ahead' : 'missing'
 }
 
 /**
@@ -121,11 +129,12 @@ export function collectDeps(
       if (name === '' || seen.has(name)) continue
       seen.add(name)
 
-      states.push({
-        spec,
-        name,
-        state: isSatisfied(installed[name], pin) ? 'present' : 'missing',
-      })
+      const state = compareSpec(installed[name], pin)
+      states.push(
+        state === 'ahead'
+          ? { spec, name, state, installed: installed[name] }
+          : { spec, name, state },
+      )
     }
   }
 

package/tooling/web/reference.md

@@ -45,6 +45,8 @@ The web layer covers the tooling shared across Vite + React, Astro, and Next.js.
 
 ## Screenshot cases
 
+The starter `e2e/home.spec.ts` is seeded once. Sync copies it into a project that has none and leaves it alone after, so grow it freely.
+
 - Edit the `CASES` in the seeded `e2e/capture/cases.ts`, one entry per output file naming a section, a theme, and a route, since sync overwrites the `e2e/screenshot.ts` harness
 - Move cases from `e2e/screenshot.ts` to `e2e/capture/cases.ts` before a sync, and set `EVIDENCE_ROOT = 'evidence'` to keep an existing baseline
 - Expect each case crossed with `VIEWPORTS`, written to `screenshots/<hostname>/<section>/<width>--<theme>.png`, so local and deployed runs stay apart
@@ -76,7 +78,7 @@ The web layer covers the tooling shared across Vite + React, Astro, and Next.js.
 
 - Expect the sweep under `screenshots/` to be ignored and only a flagged case's `EVIDENCE_ROOT` output tracked, so a target's first capture is the baseline it commits
 - Size a case's `sections` list against the frames each later change regenerates, since a case naming sections commits a folder per section
-- Run the capture after a surface changes, component files included, per `governance/rules/ui/surface-capture.md`, unless the production build strips the component out
+- Run the capture after a surface changes, component files included, per the `ui-craft` skill's capture reference, unless the production build strips the component out
 
 ## Screenshot scripts and workflows
 

package/governance/ledger/ui/copy.md

@@ -1,49 +0,0 @@
----
-title: UI copy rule ledger
-description: Sources for governance/rules/ui/copy.md, one row per bullet
----
-
-# UI copy rule ledger
-
-The field is content design, and the authority for researched interface copy is the GOV.UK Design System and its style guide.
-
-## Punctuation
-
-| Bullet                        | Source                                                                                                                                                                  | Read       |
-| ----------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------- |
-| No em dash or semicolon       | House choice: both read as generated copy in an interface, and a comma, a period, or a split sentence carries the same meaning in a shorter line                        | 2026-10-18 |
-| No trailing label punctuation | [Button](https://design-system.service.gov.uk/components/button/): every example label, such as "Start now", "Sign in", and "Continue", carries no trailing punctuation | 2026-10-18 |
-
-## Button copy
-
-| Bullet            | Source                                                                                                                                     | Read       |
-| ----------------- | ------------------------------------------------------------------------------------------------------------------------------------------ | ---------- |
-| The action's verb | [Button](https://design-system.service.gov.uk/components/button/): "Write button text in sentence case, describing the action it performs" | 2026-10-18 |
-
-## Error messages
-
-| Bullet            | Source                                                                                                                                                                                                                                            | Read       |
-| ----------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------- |
-| Say how to fix it | [Error message](https://design-system.service.gov.uk/components/error-message/): "Describe what has happened and tell them how to fix it", and "'Enter your first name' is clearer, more direct and natural than 'First name must have an entry'" | 2026-10-18 |
-| Short and plain   | [Error message](https://design-system.service.gov.uk/components/error-message/): "The message must be in plain English, use positive language and get to the point", with no "technical jargon like 'form post error'"                            | 2026-10-18 |
-
-## Placeholder text
-
-| Bullet                         | Source                                                                                                                                                                                                                                           | Read       |
-| ------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ---------- |
-| No placeholder, a hint instead | [Text input](https://design-system.service.gov.uk/components/text-input/): "Do not use placeholder text in place of a label, or for hints or examples", since "it vanishes when the user starts typing" and "not all screen readers read it out" | 2026-10-18 |
-
-## Declined
-
-Every bullet the rewrite removed or reshaped, grouped by why.
-
-### Contradicted by the source
-
-- Replaced "show format examples over instructions in placeholder text" with the hint text bullet, since GOV.UK says not to use a placeholder "for hints or examples".
-- Cut "reserve instructions for labels or helper text", which the new placeholder bullet now carries.
-
-### Merged or reworded
-
-- Merged the imperative verb and padded phrasing bullets. The `Save` over `Save changes` example went, since GOV.UK adds words "to better describe the action".
-- Merged "keep messages short and actionable" and "one sentence maximum" into the plain language bullet.
-- Reworded the error example to `Enter an email address`, the instruction form GOV.UK prefers over a description.