- The mouse cursor and the blocker-list icons are Phosphor files (media/teaser/icons, copied unchanged by
tools/site-icons.ts) shown as CSS masks; no hand-written SVG left in media/teaser.
- The no-handmade-svg guard now covers media/teaser: no inline SVG, every referenced icon is a vendored Phosphor file,
byte-equal to the package.
- New 6-bar desktop scene before the Submit card: real frames of apps/desktop in fixture mode (media/teaser/desktop.ts,
captured headlessly) in a CSS window: portfolio across studios, the release track of both stores, the Growth chart
with its tooltip, the Cmd+K palette, a price-sheet dry-run diff, the confirm dialog, and Submit for review left to
the console. Music and timeline follow (26 bars, 48.75 s).
- End card: "Desktop app for macOS, Windows, Linux".
- site/img/teaser.mp4 (1280x720 H.264, about 3 MB) and its poster re-rendered; ?v= bumped, captions say 49 s.
Desktop icons move from lucide to Phosphor (@phosphor-icons/react, per-icon imports, regular weight; fill for the
active nav item, notices and rating stars). The site's icons and the fixture art use Phosphor's own files from
@phosphor-icons/core. App Store and Google Play are shown with their marks from simple-icons (CC0 data), vendored
unchanged by tools/site-icons.ts and rendered as a mask in the neutral text colour next to the store's name; the
trademark note is in docs/desktop.md and the site footer. lucide-react and lucide-static removed. The
no-hand-written-SVG guard allows only the vendored files (byte-equal to their packages) and per-icon Phosphor
imports. Site stylesheet ?v= bumped.
- desktop: Growth chart, Health and portfolio sparklines and the rollout bar drawn with visx (responsive width,
crosshair tooltip with real values, arrow keys, theme tokens, no animation); icons from lucide-react behind the
existing Icon names; empty states are lucide icons on token tiles; store marks are generic lucide icons; stars are
lucide; the brand mark moves to ui/brand.tsx as the one allowed inline SVG
- fixtures: app icons and store art stand-ins are PNGs rendered by tools/fixture-art.ts (lucide on gradients)
- site: icons are lucide's own files in site/img/icons (tools/site-icons.ts, with the ISC licence) shown as CSS masks;
the journey diagram is rendered from docs/diagrams/journey.mmd by tools/diagrams.ts (Mermaid CLI), with a tall
variant for phones
- tests/no-handmade-svg.test.ts fails on new inline SVG drawing in apps/desktop/src and site/*.html
- desktop bundle: 598.7 -> 695.5 kB minified, 174.6 -> 206.0 kB gzip
A live run where only the token exchange, edit insert / delete and Reporting :query were let through:
- health / validate: errorIssues:search answered 400 "Unsupported timezone" for America/Los_Angeles (it takes
hour-aligned UTC only); the failure dropped the crash and ANR rates and the reason was cut off in the query string.
Issues are asked in UTC; a failing issues search is a vitals-issues warning; Reporting errors leave out the query.
- play listing --dry / play images --dry wrote into the edit (PUT details and listings, image clears and uploads,
:validate) before discarding it. Now one throwaway edit is read (details, listings, image sha256) and compared
with the local copy, then deleted.
- play.read(): one helper for read-only edits (status, next-version-code), always deleted, warns if the delete fails;
next-version-code counts distinct codes.
- tests on sanitized recorded answers (tests/play-post-reads.test.ts)
prices export --json reads the current prices without writing; inspect --declarations gives Data safety, the App
Privacy answers it suggests, app info, export compliance and App Review details with every question only the owner
can answer; play data-safety keeps the uploaded CSV in .storebound/ so changes since the upload show (Play cannot read
the form back); shots gallery --json carries the renders folder; inspect --render returns one render as a data: URL
from inside the renders folder only.
storebound import fastlane [--dir] [--out] [--profile] [--dry] (MCP import_fastlane, dry run by default) reads
Appfile / Deliverfile plain option lines, deliver and supply metadata and screenshots, and writes
storebound.config.ts, store/appstore.md, store/listing.md, the Play icon and feature graphics, and frames with a
folder capture source. App Store and Play language codes are paired (cli/lib/store-locales.ts). Existing files are
never overwritten (.imported copies with a diff hint); keys, team ids, the review contact and demo account are only
named in the report. Fixture tests/fixtures/fastlane-sample, skill migrate-from-fastlane, README and CHANGELOG.
- API error bodies on one line (errorBody): Google and App Store Connect pretty-print errors, which hid
SERVICE_DISABLED from the Play account check (reported permission instead of api-disabled), broke tables
and turned serve's apps.list error into a JSON fragment
- apps list: failure says the reason and the fix (same words as apps check)
- newest App Store version by createdDate, iOS first (the API groups versions by platform); all versions
read so no platform is missed; per-app reads four at a time
- apps check (apple): team app count from meta.paging.total with a one-app read
- Play Money without units (prices under 1.00) read as 0.xx, not NaN (catalog diff, subscription diffs)
- health: App Store Connect 404 for perfPowerMetrics / diagnosticSignatures means no field data yet;
Play vitals behind a disabled Reporting API say so with the fix
- validate: gracRatingClassificationNumber optional, socialMediaAgeRestricted only when socialMedia is yes,
socialMedia unanswered is a warning
- tests with sanitized recorded responses (tests/store-real-shapes.test.ts)
- inspect reports platforms, subscription ids and the catalog each store should sell; app identity
and parity follow app.platforms and per-item platforms; catalog.diff joins the JSON methods
- Portfolio: one Platforms column, compact catalog, paths truncated and shown with ~; the confirm
dialog and run logs show ~ instead of the home folder
- portfolio.status returns apps: identity with both platforms, offline config check, release trains
(current = highest version live on a store, next = highest above it; per-store phase, build,
rollout fraction) and catalog parity (config vs both stores) when live
- inspect --json adds product ids and the subscription count
- system.tools: ffmpeg, ffprobe, Chrome / Chrome for Testing, Xcode tools, simctl, adb, git with
versions and source; STOREBOUND_TOOL_PATH (front of PATH for every command) and STOREBOUND_CHROME
(Playwright executablePath) let the desktop app hand over its own copies
apple ppo status | create --variant b[,c] | upload --variant b | start | stop: v2 experiments (platform IOS, traffic
proportion), one treatment per shots variant (max 3), a treatment localization per App Store locale, and the variant
renders synced into the treatment screenshot sets (shared MD5-idempotent set sync with apple screenshots, which no
longer creates sets on --dry). start / stop ask first and are not MCP tools; MCP: apple_ppo_status, apple_ppo_create,
apple_ppo_upload. Play experiments stay console-only (documented).
apple.iphoneMedium (off by default): iphone-medium size key, 1206 x 2622 (1179 x 2556 accepted), uploaded as
APP_IPHONE_61; rendered, checked, required and uploaded only when enabled. Apple's scaling rule documented in the config
and the store-screenshots skill.
Dual-era server: requests that carry io.modelcontextprotocol/protocolVersion are served statelessly (server/discover
with supportedVersions, capabilities, instructions, ttlMs / cacheScope; resultType complete and serverInfo in every
result's _meta; cacheable tools/list; -32022 UnsupportedProtocolVersion with data { supported, requested }; -32602 when
clientCapabilities is missing). initialize / ping / unannotated requests keep the 2025-11-25 (and older) behaviour.
Tests for both handshakes, including one stdio process serving a modern probe and a legacy client.
App Store Connect Analytics Reports API (ONGOING request, Discovery and Engagement + Downloads Standard reports,
daily instances, md5-checked gzip segments, newest processing per day) and the Play statistics exports in the
pubsite_prod_rev_ Cloud Storage bucket (UTF-16 CSV, store listing visitors / acquisitions, daily user installs;
play.reportsBucket). Snapshots in .storebound/analytics/<date>.json; compare lines them up with the git history of the
store files (7 days before vs after each change). MCP: analytics_snapshot, analytics_compare, analytics_setup.
export reads the current App Store manual (and with --with-automatic the equalized) prices and the Play purchase
option regional prices into one CSV. import parses and checks the sheet, prints a diff table (change %), and after a
confirmation writes one App Store price schedule per changed product (base territory + every manual price, future-dated
changes via start_date, price point ids looked up per territory) and the Play regional configs (CSV over current,
convertRegionPrices for regions without a price, --reconvert). MCP: prices_diff (read-only), prices_export; applying
stays a terminal command. store-iap-setup skill: price sheet section.
App Store customerReviews / customerReviewResponses and Google Play reviews.list / reviews.reply. draft writes one
Markdown file per unanswered review to .storebound/replies (front matter approved: false, never overwritten); send
only sends approved, unsent drafts that pass the checks (Play 350 characters, no incentives or rating requests,
personal data / link / tone warnings) after a confirmation, and marks them sent. MCP: reviews_list, reviews_draft;
send is never exposed.
Play Developer Reporting API v1beta1 (crashRateMetricSet / anrRateMetricSet DAILY queries in America/Los_Angeles,
user-perceived 28-day rates, worst device models, errorIssues:search) and App Store Connect perfPowerMetrics,
diagnostic signatures and TestFlight crash submissions. Thresholds 1.09 % crash / 0.47 % ANR / 8 % per device
(developer.android.com/topic/performance/vitals): over = blocker, from 80 % = warn, in validate too (--no-health).
Weekly Markdown report with week-over-week change in .storebound/health/. MCP: health (read-only), health_report.
Beta groups from apple.testflight.groups (created when missing, never changed), testers from a CSV in .storebound/
(masked in logs, confirmation before Apple emails invitations, --remove / --invite), What to Test per locale from the
App Store listing (whatToTest or whatsNew), Test Information and Beta App Review details, builds by exact number to
groups, Beta App Review submission behind --submit-review and a confirmation. Pure planners in cli/lib/testflight.ts,
shared CSV / table / JSON:API helpers. MCP: apple_testflight_status, _groups, _what_to_test, _info, _distribute
(internal groups only). Skill testflight.
- tools/gen-adapters.ts writes the Claude Code plugin + .mcp.json, Codex plugin, Gemini CLI
extension + GEMINI.md, AGENTS.md, Copilot instructions and a Cursor rule from skills/ and
docs/agents-core.md; --check fails on stale files
- tests for adapters, MCP tools and JSON-RPC, validate rules, ASO, release notes, privacy
manifest, release helpers, install merging, scrub
- bun run ci: adapters check, tests, typecheck, secrets / personal data scrub
- 0.2.0
