docs: weekly backlog re-verification and MODULE_AUDIT refresh (September 6, 2026)

[cursor]

Posted by an AI agent on kve's behalf.

Summary

• Re-verified every open item in all 12 TODOs.md files against current source, after the week's 4 PRs (Add configurable next-launch demo mode #301 demo next-launch mode, feat(WhereUI): present card estimates as visa stickers #302 visa-sticker estimates, fix(SnapshotKitTesting): make async capture regression deterministic #303 deterministic async-capture test, feat(WhereUI): edit selected regions directly in Settings #305 Settings region editor).
• Nothing closed outright; two items narrowed. PR feat(WhereUI): edit selected regions directly in Settings #305 gave RegionsSettingsView a SnapshotProviding conformance and a 10-reference suite, shrinking the Settings image-coverage item from five screens to four. The scrolledForYear half of a namesake-test item is obsolete — the symbol no longer exists — so that item narrows to LocationNamer.
• Three small P2s filed from the window's new surface: DeveloperDemoLaunchSheet snapshots a scrolling Form at a fixed .phoneLightDark frame against the full-content rule; PR feat(WhereUI): present card estimates as visa stickers #302 orphaned WhereFormat.locationCardEstimatedDays (helper + catalog key + test, no production caller); the OnboardingGate doc comment still says the gate is "Rooted at the trunk's head" after PR Add configurable next-launch demo mode #301 prepended the demo step.
• Corrected five of the August 30 audit's numbers that were wrong at its own date: WhereUI was 276 sources / 102 tests (published 274/101); totals were 695/368 (published 694/368, table summing to 693); test bundles were 25, not "26, unchanged" (PR refactor: remove empty StuffCore module #300's StuffCore removal had already landed); the settle-floor split is 39 addressable configurations, not 37 (both prior enumerations missed the Ranking Animation Lab's 2); and the Method line carried a duplicated sentence. Also fixed a WhereShortcuts count stale since PR refactor(Where): retire AI recent activity summaries #230 (four registered shortcuts, not five) and two doc citations pointing past the end of rewritten files.
• Rewrote MODULE_AUDIT.md from the refreshed backlog (700 source / 369 test / 49 image-snapshot files; 484 references across 49 suites; 25 bundles) and refreshed the two dated reference counts in SnapshotKitTesting/AGENTS.md (472 → 484), with ./sync-agents run.

Why

The weekly todo-triage pass keeps the backlog honest: it closes what shipped, corrects citations the week's diffs moved, files what the new surface shows, and regenerates the derived audit so the report never carries claims the tree contradicts. This week's lead finding is about the audit itself — five published counts were wrong at publication, all undercounts or stale carries in figures whose text claimed freshness — so the affected items now record their derivation procedure, not just the number.

Review focus

• The four-screens rewrite of the Settings SnapshotProviding item in Where/TODOs.md (including the WhereFlyoverView nuance: it is image-pinned by a suite that hosts it directly, without a conformance).
• The 37 → 39 correction in Shared/SnapshotKitTesting/TODOs.md — the Ranking Animation Lab's settledAtLeast(1.0) × 2 configurations predates the August 30 audit, so both prior "re-derived and unchanged" claims were wrong.
• The three new P2s in Where/TODOs.md, each verified against source before filing.
• The self-correction table at the bottom of MODULE_AUDIT.md ("as published / as corrected" column).

⚠️ Breaking changes

None. Documentation and backlog files only; no production code changed.

Compatibility

Not applicable — no code, build, or CI changes.

Testing

• Executed on the Linux agent: ./swiftformat --lint (0/1125), ./shellcheck (silent), ./attribution --check (up to date, 12 credits), ./snapshot-shards check (plan valid: 49 suites, 13/15/18/3 — the window's two new suites on the intake shard as designed), Python tool tests (64 tests, 1 known macOS-only failure), Ruby tool tests (75 runs, the known 12+1 platform failures). The retained-tool failures match the filed root P1 exactly; no new platform assumption appeared.
• Six read-only verification subagents re-checked every open item with file:line evidence; every changed verdict and new finding was spot-checked by hand before publishing. False alarms rejected rather than filed (recorded in the audit): the demo-mode UserDefaults latch as "a mark on the device" (it is the Inspector pattern, one-shot, DEBUG-only); DemoDataBuilder's convenience init(now:calendar:) as a Core parameter-default violation (an explicit overload on demo fixture code, not a store API); DeveloperDemoModeRow / LocationCardEstimateSticker bare #Previews as coverage debt (both pinned transitively); RegionsSettings' unlabeled .combine (its symbol is .accessibilityHidden, children read as a sentence).
• Skipped (unavailable on the Linux VM, per root AGENTS.md): ./test, the architecture lint, ./xcstrings --lint, ./sf-symbols --lint, and any simulator or pixel verification. Documentation-only change, so no test tier applies.

  

---

Note

Low Risk
Markdown and backlog edits only; no production code, build, or CI behavior changes.

Overview
Documentation-only weekly todo-triage: re-verifies open backlog items against source after four landed PRs (#301–#305) and replaces MODULE_AUDIT.md for September 6, 2026 with updated repo counts (700/369 sources/tests, 484 snapshot references, 49 suites, 25 test bundles).

The backlog itself moves modestly: no items closed, but the Settings image-coverage item narrows from five to four screens (PR #305 added RegionsSettingsView snapshots) and the namesake-test item drops obsolete scrolledForYear, leaving LocationNamer. Three new Where P2s file convention/doc debt from the week (demo launch sheet fixed-frame Form snapshot, orphaned WhereFormat.locationCardEstimatedDays after the sticker redesign, stale OnboardingGate “trunk head” comment).

Across TODOs.md files and audit text, the pass corrects stale citations and counts (e.g. settle-floor 37 → 39, WhereShortcuts five → four, CreditKit generator path, Broadway Project.swift lines) and documents five wrong numbers in the prior audit edition. SnapshotKitTesting/AGENTS.md reference totals update 472 → 484.

^{Reviewed by Cursor Bugbot for commit a1d7246. Bugbot is set up for automated code reviews on this repo. Configure here.}