docs: record what phase 8 built, and what is left for a person across all nine

The plan's last phase becomes a record, and the document gains a closing status:
every count the audit was written to move, from the state in "Where this app
stands" to what `m3-audit.sh` reports today, and a gathered list of what a person
still has to look at — the eight screens with competing filled buttons, the two
list-detail families the pane work did not reach, the container transform, desktop
keyboard traversal, and the avatar picker's selected state.

**The audit caught the previous commit.** `ThemeGallery` added eight string
literals in composables, taking the count 39 -> 47, which is exactly the drift the
budget exists to notice. They are sample text — the words are chosen to be words,
so that colour pairings can be looked at — and putting them in the catalogue would
add eight entries no screen shows and a translator would have to be told to
ignore.

So the audit grows a third exemption marker beside `m3-color-exempt` and
`m3-spacing-exempt`: `m3-string-exempt`, per file rather than per line, because
the exemption is a property of what the file is for and eight markers down one
gallery would say less than one at the top of it. Back to 39, and the report now
says how many files are exempt so the mechanism cannot be used quietly.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

CLAUDE.md

@@ -73,6 +73,13 @@ Compose Resources is not aapt: it does *not* unescape `\'` and does *not* collap
 though it does process `\n`. `StringCatalogueJvmTest` asserts each escape the app depends
 on; add to it rather than assuming a family rule.
 
+A file whose strings are sample text rather than UI text — a gallery of colour pairings,
+say — marks itself once at the top:
+
+```kotlin
+// m3-string-exempt: these words are sample text for looking at colour pairings
+```
+
 **Budget: 0 title-case strings.** Literals in composables are reported without a budget —
 39 remain, all of them terms of a `+` concatenation.
 

composeApp/src/commonMain/kotlin/press/mantra/compose/ui/theme/ThemeGallery.kt

@@ -1,3 +1,7 @@
+// m3-string-exempt: the words in this file are sample text for looking at colour
+// pairings, not UI text. Nobody navigates here and nothing here is translated;
+// putting these eight strings in the catalogue would add eight entries that no
+// screen shows and that a translator would have to be told to ignore.
 package press.mantra.compose.ui.theme
 
 import androidx.compose.foundation.layout.Arrangement

docs/material-design-conformance.md

@@ -973,25 +973,113 @@ now would mean writing it twice.
 **Why last.** These lock in what the previous phases achieved. Written earlier,
 they would only fail.
 
-**Work.**
+**Built.** Three commits over the four items: the second needed nothing, and
+finding that out is most of what it was worth.
 
-1. **CI runs the Phase 0 audit** and fails on regression: no new `.dp` literals
-   outside the theme, no new hardcoded `Color`, no new string literal in a
-   composable, no new bare `.clickable`.
-2. **The contrast test covers every scheme and every call-site pairing**, and is
-   part of the normal test run. It is the only one of these that catches a bug a
-   human reviewer reliably misses.
-3. **Preview coverage per screen**: light, dark, high-contrast, 200% font scale,
-   compact and expanded. There are already 51 previews to build on.
-4. **A short conventions note in `CLAUDE.md`** — spacing comes from the scale,
-   colour from the role, text from resources, targets are 48dp — so the rules
-   are visible at the point of writing new code rather than at review.
+1. **`:composeApp:m3Audit`, wired into `check`.** The audit has existed since
+   phase 0 and has been run by hand at the end of every phase since, which is the
+   arrangement it was written to end: a budget nobody checks at the moment the
+   number moves is a number that drifts. It shells out to
+   `docs/scripts/m3-audit.sh --check` and fails the build on a budget exceeded or
+   a floor undercut, declares the script and the ui source tree as inputs so it is
+   up-to-date-able, and skips loudly rather than failing where there is no bash.
 
-**Done when** the audit and the contrast test both run in CI, and a change
-violating any of the four rules fails.
+   Verified to bite: one `Color(0xFFAABBCC)` added to `LoadingScreen.kt` reports
+   `hardcoded Color outside theme/ 1 over budget 0` and takes the build with it.
+
+   **A Gitea Actions workflow** beside it, since the remote is a Gitea 1.25
+   instance. Two jobs on purpose: `budgets` is grep over the source tree with no
+   gradle, no SDK, no submodules and no network — which is why the audit is a
+   shell script rather than a gradle plugin — and `tests` needs the whole
+   composite chain and a cold cross-compile of secp256k1, so it is split out for a
+   runner that has the capacity. **The workflow is unverified**: this repository
+   has had no CI of any kind, so there is no runner to try it against. The gradle
+   task is the half that is proven, and it is the half that runs on every
+   developer machine regardless.
+
+2. **The contrast test was already there.** Phases 0 and 3 built it out to ten
+   assertions over all six schemes — every content role on its container, every
+   tonal surface, the twelve fixed roles, the extended brand families, the
+   composited translucent containers, `outline` at 3:1 — and it runs in
+   `:composeApp:jvmTest`, which is now a CI job. Nothing was added, and the reason
+   is recorded in the test itself: a call site that pairs two roles the scheme
+   already covers produces a pairing the first assertion already walks, so
+   restating it per site would double the maintenance and catch nothing.
+
+3. **`@ConformancePreviews` on all 53 previews.** Every one of them rendered a
+   light theme at 100% text at whatever width the pane happened to be — the only
+   condition under which this app has never had a defect. They now render under
+   five: light, dark, 200% text, compact 400dp, expanded 1000dp. Each of the four
+   new ones is where a defect has actually been. It also gives phase 6 the check
+   it could not make: two of its five acceptance widths are now one click away on
+   every screen.
+
+   High contrast is deliberately not in the annotation. Contrast is a property of
+   the scheme rather than of a screen, `ColorSchemeContrastTest` measures every
+   pair in all six, and there is no `@Preview` parameter for it — it needs
+   `TorchTheme(contrast = …)` in the body. `ThemeGallery` covers the six once,
+   over components rather than screens, with `dynamicColor = false` because an
+   android 12+ preview would otherwise paint all six columns from the wallpaper.
+   It is the only place the medium and high contrast schemes can be seen at all.
+
+4. **`CLAUDE.md`**, which this repository did not have. Seven rules, each with the
+   shape to copy, the shape not to, and the budget the audit holds it to. Where a
+   rule has a trap that has already caught somebody, the trap is named rather than
+   the rule restated — `.copy(alpha = …)` on a content role is how nine contrast
+   failures got in, Compose Resources unescapes `\n` but not `\'`,
+   `AnimatedContent` takes a `when`'s branches out of `ColumnScope`,
+   `MotionSchemeKeyTokens` is internal.
 
 **Risk:** none to the app; some friction for contributors, which is the point.
 
+---
+
+## Where this leaves the app
+
+All nine phases are built. The counts the audit was written to move, from the
+state recorded in "Where this app stands" to the state `docs/scripts/m3-audit.sh`
+reports today:
+
+| area | was | is | budget |
+|---|---|---|---|
+| roles falling to the baseline palette | 12 | 0 | 0 |
+| hardcoded `Color` outside `theme/` | 11 | 0 | 0 |
+| dp literals in spacing positions | 527 | 0 | 0 |
+| `.clickable` with no minimum target | 33 | 0 | 0 |
+| untriaged `contentDescription = null` | 18 | 0 | 0 |
+| title case in UI strings | ~45 | 0 | 0 |
+| string literals in composables | 334 | 39 | reported |
+| `stringResource` call sites | 2 | 422 | — |
+| snackbar hosts | 0 | 137 | — |
+| adaptive API uses | 2 | 12 | floor 12 |
+| navigation components | 0 | 2 | floor 2 |
+| motion API uses | 1 | 13 | — |
+| navigation transitions | 0 | 3 | — |
+
+The 39 remaining literals are all terms of a `+` concatenation, several of them
+pluralisations that want a real plural resource rather than a format argument;
+`m3-extract-formatted.py --remaining` lists them.
+
+What a person still has to look at, gathered from the phases that said so:
+
+- **which of a competing pair of filled buttons is primary**, on eight screens.
+  That is a product decision about what each screen is for, and getting it wrong
+  quietly weights a choice the user is supposed to make freely (phase 5).
+- **whether a 480dp column of a particular screen reads well.** The measure is
+  applied at every root and asserted at five widths, but "renders correctly" is a
+  judgement no assertion makes (phase 6).
+- **proposals → signing and artifacts → chapters**, the two list-detail families
+  the pane work did not reach. Same shape as chat, different content (phase 6).
+- **the container transform** between a list item and its detail screen, which
+  wants `SharedTransitionLayout` and only applies below the expanded breakpoint,
+  where the detail is not already beside the list (phase 7).
+- **full keyboard traversal on desktop** — tab order across the modal sheets and
+  the dialog, and focus returning to what opened them. Compose restores focus by
+  default, so the gap is evidence rather than known breakage (phase 3).
+- **the avatar picker's selected state**, `secondaryContainer` at 1.65:1 against
+  the surface, which M3 accepts only where a second cue carries the selection.
+  This grid has neither an outline nor a checkmark (phase 3).
+
 ## Order and dependencies
 
 | phase | depends on | touches | reversible alone |

docs/scripts/m3-audit.sh

@@ -209,8 +209,32 @@ note "TextAlign.Center:                       $centred"
 # ---------------------------------------------------------------------------
 hdr 'Content  (phase 4)'
 
-literals=$(( $(count 'text = "') + $(count 'Text\("') ))
+# A file whose strings are sample text rather than UI text -- ThemeGallery, whose whole
+# job is to render colour pairings and whose words are chosen to be words -- marks itself
+# once at the top with `// m3-string-exempt: <reason>`. Per file rather than per line,
+# because the exemption is a property of what the file is for, and eight markers down one
+# gallery would say less than one at the top of it.
+#
+# The same shape as `m3-color-exempt` and `m3-spacing-exempt`: the reason travels with the
+# code rather than living in a list of file names here.
+sample_text_files() {
+  grep -rlE 'm3-string-exempt' "$UI" --include=*.kt 2>/dev/null
+}
+
+string_literals() {
+  local exempt
+  exempt=$(sample_text_files | sed 's|^|^|' )
+  if [[ -z $exempt ]]; then
+    grep -rE 'text = "|Text\("' "$UI" --include=*.kt 2>/dev/null | grep -vE "$NOT_A_COMMENT"
+  else
+    grep -rE 'text = "|Text\("' "$UI" --include=*.kt 2>/dev/null | grep -vE "$NOT_A_COMMENT" \
+      | grep -vE "$(sample_text_files | paste -sd'|' -)"
+  fi
+}
+
+literals=$(string_literals | wc -l | tr -d ' ')
 report 'string literals in composables' "$literals" "$BUDGET_STRING_LITERALS"
+note "sample-text files exempt: $(sample_text_files | wc -l | tr -d ' ')"
 res=$(count 'stringResource|Res\.string')
 note "stringResource / Res.string:            $res"
 # Delegated, because the grep version was wrong twice: it required every word after the