Install the "release-prepare" agent skill from https://github.com/n0an/VivaDicta/tree/main/.agents/skills/release-prepare into .claude/skills/release-prepare/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "release-prepare", then confirm the skill loads.
Claude Code copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
Type this inside Codex. $skill-installer <name> installs a curated skill from openai/skills. The installer writes to $CODEX_HOME/skills (default ~/.codex/skills). Restart Codex if the skill does not show up.
skills CLI
$ npx skills add n0an/VivaDicta --skill release-prepare -a codex
Project install goes to .agents/skills/; add -g for ~/.codex/skills/.
Install the "release-prepare" agent skill from https://github.com/n0an/VivaDicta/tree/main/.agents/skills/release-prepare into .agents/skills/release-prepare/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "release-prepare", then confirm the skill loads.
Codex copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
skills CLI
$ npx skills add n0an/VivaDicta --skill release-prepare -a cursor
Project install goes to .agents/skills/; add -g for ~/.cursor/skills/.
Install the "release-prepare" agent skill from https://github.com/n0an/VivaDicta/tree/main/.agents/skills/release-prepare into .cursor/skills/release-prepare/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "release-prepare", then confirm the skill loads.
Cursor copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
--scope user (default) or --scope workspace; --path is the subfolder of the repo that holds the skill; --consent skips the security confirmation prompt.
skills CLI
$ npx skills add n0an/VivaDicta --skill release-prepare -a gemini-cli
Project install goes to .agents/skills/; add -g for ~/.gemini/skills/.
Install the "release-prepare" agent skill from https://github.com/n0an/VivaDicta/tree/main/.agents/skills/release-prepare into .gemini/skills/release-prepare/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "release-prepare", then confirm the skill loads.
Gemini CLI copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
GitHub CLI
$ gh skill install n0an/VivaDicta release-prepare
Installs for Copilot at project scope by default; add --scope user for a personal install. Preview a skill first with gh skill preview. Needs GitHub CLI 2.90.0 or later (public preview).
skills CLI
$ npx skills add n0an/VivaDicta --skill release-prepare -a github-copilot
Project install goes to .agents/skills/; add -g for ~/.copilot/skills/.
Install the "release-prepare" agent skill from https://github.com/n0an/VivaDicta/tree/main/.agents/skills/release-prepare into .github/skills/release-prepare/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "release-prepare", then confirm the skill loads.
GitHub Copilot copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
skills CLI
$ npx skills add n0an/VivaDicta --skill release-prepare -a opencode
OpenCode documents no install command of its own. Project install goes to .agents/skills/; add -g for ~/.config/opencode/skills/.
Install the "release-prepare" agent skill from https://github.com/n0an/VivaDicta/tree/main/.agents/skills/release-prepare into .opencode/skills/release-prepare/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "release-prepare", then confirm the skill loads.
OpenCode copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
Facts
Skill name
release-prepare
GitHub stars
131
Token cost
~7.8k tokens
SKILL.md length
3,553 words
Files
3 (incl. references)
Skills in repo
20
Repo updated
First seen
Licence
MIT
At a glance
Prepare a new VivaDicta release — version bump, What's New screen, App Store metadata, code sweep, and pre-submission checklist
Works in 12 steps: Create release branch → Bump version numbers → Code sweep → …
Tasks that involve App store release
SKILL.md covers Version Format Convention, Related Skills and Skill Flow
Calls git, gh and npm; reaches vivadicta.com
What it does
Release Prepare is an agent skill from n0an/VivaDicta. Prepare a new VivaDicta release — version bump, What's New screen, App Store metadata, code sweep, and pre-submission checklist
Its SKILL.md is about 7.8k tokens, which your agent loads only when the skill is triggered. The skill folder holds 4 other files, including reference files (for example `agents/openai.yaml` and `references/whats-new-screen.md`).
It sits in Mobile, covering App store release. The repository describes itself as: iOS & watchOS speech-to-text app with AI voice keyboard, on-device RAG, and chat with your notes - powered by Apple Foundation Models, WhisperKit, NVIDIA Parakeet, and 20+ AI…. The licence is MIT.
When your agent uses it
Tasks that involve App store release
Example prompts
“/release-prepare”
Requirements
Python 3
Workflow steps
12 steps, taken from the step headings in SKILL.md.
Read from SKILL.md and the folder at commit b141304. It shows what the files ask for, not the result of running them.
Tool permissions
Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.
From allowed-tools in the SKILL.md frontmatter.
Runs code
Shell commands in SKILL.md call:
git
gh
npm
xcodebuild
python3
From the folder's file list and the shell code blocks in SKILL.md.
Network
Hosts in commands or code, which the agent is likely to contact:
vivadicta.com
Also links to:
icloud.developer.apple.com
From URLs in SKILL.md, links to its own repository left out.
Credentials
Names no API keys, tokens, secrets or passwords.
From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.
Context cost
Release Prepare loads about 7.8k tokens when it runs, and up to ~8.7k if it reads all its reference files. Until then it costs about 36 tokens; SKILL.md has 3,553 words of instructions outside code blocks.
Always· name and description, kept in context so the agent knows when to use it
~36
When it runs· the whole SKILL.md, loaded when a task matches
~7.8k
With references· SKILL.md plus every file in references/, read only if the agent opens them
~8.7k
Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.
Safety
Auto-check passed
The automated check found no risky patterns in SKILL.md.
Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.
Download SKILL.mdSave it as .claude/skills/release-prepare/SKILL.md (or your agent's skills folder). This skill also uses 2 other files; get the full folder from GitHub.
name
release-prepare
description
Prepare a new VivaDicta release — version bump, What's New screen, App Store metadata, code sweep, and pre-submission checklist
Release Prepare
Use this skill when preparing a new VivaDicta release for App Store submission.
Version Format Convention
Use the full X.Y.Z form in every user-facing surface — never the short X.Y form like "3.3":
In-app What's New screen headline field (e.g. "What's New in VivaDicta 3.3.0")
App Store release notes body text (e.g. "VivaDicta 3.3.0 adds...")
App Store description if it mentions the version
Website changelog intro paragraph (the <h2> already shows 3.3.0, but the prose must too)
The only place short X.Y form survives is the internal WhatsNewCatalog.releases dictionary key, because the version-match logic is intentionally major.minor (so 3.3.0 / 3.3.1 / 3.3.2 all hit the same entry). That key is not user-visible.
Related Skills
asc-release-flow — drive the App Store Connect submission flow (Step 11 covers the normal archive-upload-and-submit path; reach for this skill when a submission needs unpicking)
asc-whats-new-writer — generate App Store release notes
asc-metadata-sync — sync and validate App Store metadata
asc-localize-metadata — sync metadata across localizations (used for ASO, not actual translation)
asc-aso-audit — run ASO audit on App Store metadata and surface keyword gaps
Skill Flow
Example queries:
"prepare release 2.2.0"
"let's ship a new version"
"release prep"
Step 1 — Create release branch
git checkout -b release/X.Y.Z
Step 2 — Bump version numbers
Update MARKETING_VERSION and CURRENT_PROJECT_VERSION in project.pbxproj across ALL 9 targets × 3 configs (Debug, QA, Release) = 27 entries each:
VivaDicta (main app)
VivaDictaKeyboard
VivaDictaWidgetExtension
ShareExtension
ActionExtension
VivaDictaTests
VivaDictaWatch Watch App
VivaDictaWatch Watch AppTests
VivaDictaWatchWidgetExtension
Don't hardcode the expected count - targets come and go (a VivaDictaWatch Watch AppUITests target existed when this was written, which is why the doc long said 30/10). Derive it instead:
bash
grep -c "MARKETING_VERSION = " VivaDicta.xcodeproj/project.pbxproj # current entry count
Then confirm the same number carries the new value after the sed.
Fastest approach - a global sed replace catches every entry in one shot:
bash
N=$(grep -c "MARKETING_VERSION = " VivaDicta.xcodeproj/project.pbxproj) # baseline, currently 27
grep -cE "MARKETING_VERSION = {OLD}" VivaDicta.xcodeproj/project.pbxproj # should equal $N
sed -i '' 's/MARKETING_VERSION = {OLD};/MARKETING_VERSION = {NEW};/g' VivaDicta.xcodeproj/project.pbxproj
sed -i '' 's/CURRENT_PROJECT_VERSION = {OLD_BUILD};/CURRENT_PROJECT_VERSION = {NEW_BUILD};/g' VivaDicta.xcodeproj/project.pbxproj
# verify
grep -E "MARKETING_VERSION|CURRENT_PROJECT_VERSION" VivaDicta.xcodeproj/project.pbxproj | sort -u # should show only 2 unique lines
grep -c "MARKETING_VERSION = {NEW}" VivaDicta.xcodeproj/project.pbxproj # should equal $N
Convention: CURRENT_PROJECT_VERSION is a monotonic counter, independent of MARKETING_VERSION. Bump it by +1 for every new build uploaded to TestFlight/App Store Connect, regardless of whether the marketing version changed. Apple only requires the build number to be strictly greater than any previously uploaded build for the same marketing version, so a plain incrementing integer is the simplest correct approach.
Before editing, check the current value:
grep -E "CURRENT_PROJECT_VERSION" VivaDicta.xcodeproj/project.pbxproj | head -1
Then use current + 1 as the new value across all targets.
The same number must be passed to asc publish appstore --build-number in Step 11 - local-build mode does not read it from the pbxproj.
Legacy note: earlier releases up through 3.0.0 used a packed XYZN scheme (e.g. 1.1.0 → 1101, 3.0.0 → 3001), which is why the counter currently sits at a value like 3001-ish. That scheme is abandoned because it caps each segment at 9. Going forward, just +1 from the last build number - do NOT try to re-pack based on the marketing version.
Step 3 — Code sweep
Run a pre-release code sweep checking for:
TODO/FIXME/HACK comments that need attention before release
Hardcoded debug/test values (test API keys, localhost URLs, debug flags)
print() statements that should use Logger
Dangerous force unwraps
Any leftover debug triggers (e.g., forced What's New screen)
Build the project and check for warnings
Step 4 — What's New in-app screen
Use references/whats-new-screen.md for the full guide on adding What's New content to the in-app screen.
Two things that are easy to miss, both covered in that guide:
Prefix the new release_X_Y property with an ISO date comment (// YYYY-MM-DD) - WhatsNewRelease has no date field, so this comment is the file's only record of when the entry was written. Don't backfill older entries.
The headline uses the full X.Y.Z form; only the releases dictionary key is X.Y.
Step 5 — App Store What's New (release notes)
Write App Store release notes and save to Obsidian vault at:
Projects/VivaDicta/what's new/whats-new-X.Y.Z.md
Read the guide first: Projects/VivaDicta/description/description-guide.md - covers structure, writing rules, and common pitfalls for both What's New and description text.
Format: No frontmatter, no markdown header — plain text only, ready to copy-paste into App Store Connect.
IMPORTANT: App Store What's New (release notes) limit is 4,000 characters. Always verify the character count.
ASO strategy (important, easy to misread):
description + whatsNew → identical English text in all 10 locales (no translation)
keywords → unique per locale (this is the ASO hack - different keyword sets target different search markets)
marketingUrl + supportUrl → same across locales
Primary source - the running draft: Projects/VivaDicta/what's new/whats-new-running.md is a running accumulator of user-facing What's New items for the upcoming release. Add to it as features land between releases, then use it as the main source for both the in-app What's New (Step 4) and these App Store release notes.
After the release ships, empty whats-new-running.md - clear all items, leaving only the Running What's New (next release) header line. It tracks only the next upcoming release, so its items must be cleared once they have shipped in whats-new-X.Y.Z.md. (Also in the Step 10 checklist.)
Framing note (Apple Foundation Model): when a release adds on-device / local AI, only recommend Apple Foundation Model on Apple-owned surfaces - the App Store description and these release notes - so App Review sees we are not positioning against Apple's own model. On our own channels (in-app What's New, website changelog, LinkedIn), present the on-device feature directly, without recommending Apple FM. The "works even on devices without Apple Intelligence, including iOS 18" value angle is fine everywhere - it is a capability statement, not a recommendation.
Step 6 — App Store description
Check if the current App Store description needs updating for new features.
Read the guide first: Projects/VivaDicta/description/description-guide.md - covers section ordering, core identity rules, jargon avoidance, and Apple Foundation Model placement.
Previous descriptions are stored at: Projects/VivaDicta/description/
Format: No frontmatter, no markdown header — plain text only, ready to copy-paste into App Store Connect.
IMPORTANT: App Store description limit is 4,000 characters. Always verify the character count before finalizing.
If updating, save the new version as description-X.Y.Z.md in the same directory.
Step 7 — Generate ASC metadata directory for the new version
Create metadata/version/X.Y.Z/*.json for all 10 locales, ready to push to App Store Connect via asc. The directory is gitignored - ASC is source of truth, this is just the staging payload.
Seed from the previous version, overwrite description and whatsNew with the new English text, keep per-locale keywords/marketingUrl/supportUrl untouched:
bash
mkdir -p metadata/version/{NEW}
python3 << 'PYEOF'
import json
locales = ['ar-SA','en-US','es-MX','fr-FR','ko','pt-BR','ru','vi','zh-Hans','zh-Hant']
with open("/Users/antonnovoselov/Documents/Vault/Projects/VivaDicta/description/description-{NEW}.md") as f:
new_desc = f.read().rstrip('\n') + '\n'
with open("/Users/antonnovoselov/Documents/Vault/Projects/VivaDicta/what's new/whats-new-{NEW}.md") as f:
new_wn = f.read().rstrip('\n') + '\n'
for loc in locales:
with open(f'metadata/version/{PREV}/{loc}.json') as f:
d = json.load(f)
d['description'] = new_desc
d['whatsNew'] = new_wn
with open(f'metadata/version/{NEW}/{loc}.json', 'w') as f:
json.dump(d, f, indent=2, ensure_ascii=False)
PYEOF
Push happens during submission (see Step 10):
bash
# NOTE: the subcommand is `apply`, not `push` - there is no `asc metadata push`.
asc metadata apply --app 6758147238 --version X.Y.Z --platform IOS --dir ./metadata --dry-run # always dry-run first
asc metadata apply --app 6758147238 --version X.Y.Z --platform IOS --dir ./metadata
asc validate --app 6758147238 --version X.Y.Z --platform IOS # expect 0 errors / 0 blocking; the info-level App Privacy advisory is permanent, ignore it
Step 8 — Update feature changelog
Move shipped features from "Unreleased" to "Released" section in:
Projects/VivaDicta/feature-changelog.md (Obsidian vault)
Also add a new ### vX.Y.Z (YYYY-MM-DD) section with feature groupings and PR numbers (see earlier entries for format).
Step 8.5 — Update iOS website changelog
The changelog is data-driven - do not hand-write JSX. Add a new entry to the releases array in:
/Users/antonnovoselov/Desktop/_Projects/iOS/VivaDictaMeta/vivadicta_website/components/ios-changelog/releases-data.tsx
app/ios/changelog/page.tsx renders the index from that array, and app/ios/changelog/[version]/ generates one static page per entry - so a new entry automatically creates https://vivadicta.com/ios/changelog/X.Y.Z, which is the URL the LinkedIn post links to (Step 8.6).
Insert the new object at the top of the array (newest first). Required fields:
tsx
{
version: "X.Y.Z", // also the URL slug
date: "YYYY-MM-DD", // ISO, used for sorting + sitemap lastModified
dateLabel: "Month D, YYYY", // shown in the UI
summary: "...", // lead paragraph; ReactNode, so it can carry inline <Link>s
summaryText: "...", // plain-text lead for <meta description> / Open Graph
groups: [
{ heading: "Feature Group", items: ["bullet 1", "bullet 2"] },
],
}
Gotchas:
The file is .tsx, not.ts - grep/find for releases-data.ts will miss it.
items entries containing an apostrophe or quotes need the string quoting swapped ('...' vs "...") or escaping, since they are plain JS strings.
Verify with npm run build and confirm the log lists /ios/changelog/X.Y.Z among the generated routes.
Keep copy aligned with the in-app What's New and App Store release notes so messaging stays consistent across surfaces. Commit and push in the website repo (separate from the iOS app repo - the website repo is auto-commit, no need to ask).
Step 8.6 — Draft LinkedIn announcement post
Prepare a launch post for the VivaDicta company page on LinkedIn (not Anton's personal feed - the voice is announcement-style, not first-person story).
Save two files to /Users/antonnovoselov/Documents/Vault/Projects/VivaDicta/linkedin-posts/:
linkedin-X.Y.Z-<slug>.md - the working draft (all variants + posting notes below). Normal blank lines so it reads cleanly in Obsidian.
linkedin-X.Y.Z-<slug>-copy.md - the paste-ready primary draft only, formatted for LinkedIn's composer (see "LinkedIn composer formatting" below). No headings, no frontmatter, no variants - just the exact text that goes into the composer.
The working draft (.md) should include:
Primary draft - ready to paste. Lead with a one-line release announcement, one paragraph framing the headline feature, then a tight bullet list (4-6 items) of other release highlights, then a link.
Spartan alternate - bullets only, no narrative paragraph. Mirrors the Summit AI Notes template (line 1 = announcement, then bullets, "and more.", link).
2-3 alternate hook lines - so the user can swap if the marquee feels wrong.
Posting notes - media suggestion, recommended post window (Tue/Wed/Thu, 9-11am), instruction to pin the App Store link as the first comment.
Style rules (also in ~/.claude/skills/linkedin-post-style/SKILL.md):
Company-page voice: no "I", no personal-story narrative, no "thrilled to announce" filler.
Short paragraphs (1-2 sentences), line breaks over dense blocks.
Normal dashes, never em-dashes.
For unordered lists use ➡️ as the bullet marker (one per item, followed by a space), never -, *, or •. Numbered lists still use 1., 2., etc.
The link in the body should be the version-specific website changelog page (https://vivadicta.com/ios/changelog/X.Y.Z, e.g. https://vivadicta.com/ios/changelog/3.5.0) - the per-release deep page, not the generic /ios/changelog list and not the App Store URL. Pin the App Store URL as the first comment instead - this gives commenters two routes (deep-context page in body, one-click install in comment).
For the marquee feature, use a tagline that reads as what the experience feels like, not what it historically was. Avoid hyperbolic "only [elite group] had this" framing - LinkedIn's tech-literate audience will dunk on overstatement, especially for capabilities competitors have shipped (real-time translation, on-device AI, etc.).
LinkedIn composer formatting (the -copy.md whitespace hack): LinkedIn's web composer collapses true blank lines when you paste plain text, destroying the paragraph spacing. So in the paste-ready -copy.md, every otherwise-blank line between paragraphs must contain a single Braille pattern blank character ⠀ (U+2800), not be empty. LinkedIn treats a line containing ⠀ as having content and preserves it as visible spacing, and the character renders invisibly to readers. Apply this only to -copy.md - leave the working .md draft with normal blank lines so it stays readable in Obsidian. (Same convention as the user-level linkedin-post-style skill, which is the source of truth for Anton's LinkedIn formatting.)
Publishing happens after the App Store build is approved and live, so users hitting the post can install immediately. The draft is prepared during release prep so it's ready to go when approval lands.
These are easy to forget - nothing auto-generates them, and the major chatbots do not auto-fetch /llms.txt (it's a voluntary convention; mainly agents/dev tools that deliberately fetch it benefit). Still, on each release re-sync them against this release's changes so anything that does read them sees current data:
Transcription providers - the cloud list + default models (source of truth: iOS AIProvider.swiftdefaultModel / availableModels and TranscriptionKit).
AI enhancement providers - the ## AI Enhancement Providers table, the provider count ("N+ AI providers"), and the one-line provider list in llms.txt (source: AIProvider.generalProviders). Add any provider shipped this release.
Preset count / categories, translation languages, supported file formats, pricing, system requirements - update only if they changed this release.
Keep wording aligned with the in-app What's New, App Store notes, and website changelog (Steps 4, 5, 8.5).
Commit and push in the website repo (can be folded into the same commit as Step 8.5, since both touch that repo).
Step 9 — CloudKit schema deployment
First, detect whether any SwiftData @Model classes changed since the last release tag:
Check if Indexes/Record Types/Security Roles show "Modified"
If yes → click Deploy Schema Changes... at the bottom of the sidebar → confirm → Deploy
If no SwiftData models changed, skip this step.
See Projects/VivaDicta/CloudKit Schema Deployment.md in the Obsidian vault for full details.
Show full SKILL.md (1,565 more words)Show less
Step 10 — Final checklist
Known-answer ASC questions (don't re-ask these):
App Privacy: already published and unchanged since 3.8.0. The privacy label is per-app, not per-version, so it carries forward on its own and nothing needs confirming each release. asc validate always reports privacy.publish_state.unverified at info level because publish state is not exposed by the public API - that line is expected and is not a blocker. Only revisit if a release starts collecting data it did not before, which means a new SDK or a new analytics event, not a new feature.
Export compliance / encryption: always "No - app does not use non-exempt encryption". The app doesn't ship its own encryption; any HTTPS usage is covered by the standard exemption. This is now declared permanently by ITSAppUsesNonExemptEncryption = false in VivaDicta/Info.plist, so uploaded builds arrive already exempt and nobody should be asked. If a build ever shows n/a in asc builds list, that key went missing - fix the plist rather than patching the build.
Before shipping (Step 11):
Version and build number bumped across all 9 targets (27 pbxproj entries - verify with grep -c)
What's New in-app screen content added, with the // YYYY-MM-DD comment above the new release_X_Y property
No debug triggers left in code (forced What's New, test flags, etc.)
Project builds with no errors
App Store release notes prepared (vault + whats-new-X.Y.Z.md)
App Store description updated if needed (under 4,000 chars)
metadata/version/X.Y.Z/*.json generated for all 10 locales
LinkedIn announcement drafted - working .md + paste-ready -copy.md (with ⠀ U+2800 blank lines and ➡️ bullets) in Projects/VivaDicta/linkedin-posts/ - publish after App Store approval lands
CloudKit schema deployed if SwiftData models changed
Review Notes: testing instructions only (remove any rejection-specific notes from previous submissions)
Changes committed and pushed on release branch
asc publish appstore run with an explicit --build-number (dry-run first; a Build Number of 1 means the flag is missing)
After upload: build is VALID with Encryption: exempt, and asc validate returns 0 errors / 0 blocking (the permanent info-level App Privacy advisory is expected)
After submitting: merge to main, annotated tag vX.Y.Z, and gh release create (Step 12)
After the release ships: empty whats-new-running.md (clear items, keep the Running What's New (next release) header) so it only tracks the next upcoming release
Step 11 — Ship it (create version → archive → upload → attach → validate → submit)
The agent drives the whole thing with asc, including the archive and upload. Anton does not touch Xcode. The one thing that still needs an explicit yes is the final submit.
Export compliance is not a step - ITSAppUsesNonExemptEncryption = false in VivaDicta/Info.plist means builds arrive already exempt (see Step 10).
1. Check ExportOptions.plist
ExportOptions.plist lives in the repo root and is committed. Local-build mode runs
xcodebuild -exportArchive under the hood, which cannot run without it:
Nothing in it is sensitive - the team ID is already in project.pbxproj. It should not
need editing; if it goes missing, recreate it with exactly these keys and plutil -lint it.
If a run ever shows Xcode overriding the build number during export, add
manageAppVersionAndBuildNumber = false and record that here.
--build-number is mandatory, not optional. Without it local-build mode does not read CURRENT_PROJECT_VERSION from the pbxproj - it auto-resolves and falls back to --initial-build-number, whose default is 1. The dry-run's Build Number column is how you catch this: if it reads 1 instead of the value bumped in Step 2, the flag is missing. Caught on the 3.10.0 release before it reached Apple.
Confirm the plan lists these seven steps, and that Will Submit is false:
ensure_version creates the App Store version if it does not exist, so a separate asc versions create is unnecessary. apply_metadata runs metadata apply against metadata/version/X.Y.Z/ from Step 7.
3. Run it for real
Drop --dry-run. Run it in the background - a full Release archive of the app plus 7 embedded targets, the upload, and ASC processing take a long time. Poll the output file rather than blocking. Wrap it in caffeinate -dims so the Mac does not sleep mid-run: the 3.11.1 export died with exportArchive The network connection was lost when the Mac went offline.
If the archive succeeded but a later step failed, do not re-archive. Export from the existing archive, then upload the IPA:
# same command, minus --dry-run, with --output json --pretty
The weak point is signing: automatic signing has to fetch distribution profiles for all 7 bundle IDs headlessly. If the login keychain prompts, the run stalls rather than failing cleanly. If it hangs with no output progress, that is the first thing to check.
Verify before running that signing is actually automatic and on one team - three stray DEVELOPMENT_TEAM = TDX2FNZ56U entries exist in the pbxproj but are overridden. grep alone will mislead you; xcodebuild -showBuildSettings -project VivaDicta.xcodeproj -target ActionExtension -configuration Release is authoritative and resolves to 358V8FBM3U.
The dry-run reports wouldSubmit and whether the build is alreadyAttached, so it is worth running even though step 3 already attached it.
Always ask before submitting, every time. Everything up to attaching the build is reversible; submission is not (only cancellable).
Fallback: Anton archives in Xcode
If headless signing fails and is not worth debugging mid-release, fall back to Product > Archive, then Organizer > Distribute App > App Store Connect > Upload (Upload, not Export). Then pick up from step 4 above, adding the attach that asc publish would have done:
To upload an already-exported IPA without local-build mode: asc publish appstore --ipa path/to.ipa - no ExportOptions.plist required.
Gotchas found in practice
asc publish appstore silently drops whatsNew on a brand-new version.apply_metadata
runs beforeattach_build, and ASC rejects the field until a build is attached, logging
'whatsNew' cannot be set for this version (initial releases have no What's New section). Retrying without it. once per locale. The run still exits 0 and reports attached: true, so
it looks clean. Always re-run asc metadata apply after the publish finishes, then confirm
with asc metadata pull into a scratch dir and check whatsNew is non-empty in all 10 locales.
Hit on 3.10.1; the release notes would have shipped blank.
Apple rejects <|...|> in whatsNew as markup: An attribute value has invalid characters.: What's New In This Version in <locale> must not contain markup language (for example, HTML tags).
Anything angle-bracketed reads as an HTML tag to the validator. This bit when the 3.10.1 notes
quoted Whisper's <|startoftranscript|> token literally - describe such text in words instead.
The failure names only the first locale (ar-SA), so it looks locale-specific and is not.
--build-number must be passed explicitly to asc publish appstore in local-build mode, or the build uploads as 1. See step 2.
asc metadata pushdoes not exist - the subcommand is asc metadata apply.
asc versions view shows empty Build ID / Build Version columns even when a build is correctly attached - ASC omits relationship linkage unless ?include= is passed, and the command does not. Do not read that as a missing build. asc validate is authoritative: its build.required.missing error clears the moment the build is attached.
asc metadata apply reports add for whatsNew on a fresh version (the field does not exist yet) and update for description. Seeing zero keywords rows is the signal the per-locale ASO keyword sets were preserved - if keywords ever appear in the plan, stop and investigate.
Run asc metadata applywithout--allow-deletes. With it, any locale missing locally is planned as a delete.
asc builds update --uses-non-exempt-encryption=false is only a patch for builds uploaded before the Info.plist key existed.
Do not git add -A on a release commit. The repo carries untracked files that are not part of the release (e.g. .claude/skills/prcdx); stage the changed files by name instead.
Step 12 — Merge, tag, and cut the GitHub release
Runs immediately after submitting, not after Apple approves. v3.9.0 was tagged the same
day it was submitted, and the tag records what was sent to Apple - approval does not change the
commit.
bash
git checkout main && git pull --ff-only
git merge --no-ff release/X.Y.Z -m "Merge branch 'release/X.Y.Z'"
git tag -a vX.Y.Z -m "VivaDicta X.Y.Z (build NNNN)"
git push origin main && git push origin vX.Y.Z
Conventions, all matching every previous release:
Tags are annotated (-a), never lightweight, and the message is exactly VivaDicta X.Y.Z (build NNNN).
Tags live on main, so the merge comes first.
Merge with --no-ff so the release branch stays visible in history.
An ## What's Changed heading, then one - bullet per user-facing change. Write these from
the same source as the App Store notes (Step 5) so the two agree, but they can be blunter -
the audience is technical.
An optional ### Internal section for tooling, skills and docs that shipped in the same range.
A closing **Full Changelog**: https://github.com/n0an/VivaDicta/compare/vPREV...vX.Y.Z line.
Regular dashes, never em-dashes, even though older release bodies contain them.
When the cycle went through PRs, gh appends the usual * … by @n0an in … list. When work was
committed straight to main there is nothing to append, so say so explicitly in one line
rather than leaving readers wondering where the PR list went.
Only mark --latest when this really is the newest release.
Release Prepare next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.
How to reuse the KMP Starter Template's built-in feature modules — Analytics, Remote Config, Purchases, Database, Store Reviews & Updates, Splash/Onboarding, Notifications, and Locale.
Generate high-converting App Store screenshots by analyzing your app's codebase, discovering core benefits, and creating ASO-optimized screenshot images using Nano Banana Pro.
Set territory-specific pricing for subscriptions and in-app purchases using current asc setup, pricing summary, price import, and price schedule commands.
Prepare a new VivaDicta release — version bump, What's New screen, App Store metadata, code sweep, and pre-submission checklist. Release Prepare is an agent skill from n0an/VivaDicta.
When should I use Release Prepare?
Release Prepare fits situations like: tasks that involve App store release.
How do I install Release Prepare in Claude Code?
Run `npx skills add n0an/VivaDicta --skill release-prepare -a claude-code`. Or copy the skill folder (.agents/skills/release-prepare in n0an/VivaDicta) into .claude/skills/release-prepare in your project. Claude Code loads it when a task matches its description.
How do I install Release Prepare in Codex?
Run `npx skills add n0an/VivaDicta --skill release-prepare -a codex`. Or copy the skill folder (.agents/skills/release-prepare in n0an/VivaDicta) into .agents/skills/release-prepare in your project. Codex loads it when a task matches its description.
Can I use Release Prepare in Cursor, Gemini CLI or GitHub Copilot?
Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add n0an/VivaDicta --skill release-prepare -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/release-prepare, .gemini/skills/release-prepare, .github/skills/release-prepare and .opencode/skills/release-prepare in your project.
What does Release Prepare need to run?
Going by SKILL.md and its folder, Release Prepare needs the command-line tools its instructions call (git, gh, npm, xcodebuild and python3). Our summary lists: Python 3.
Does Release Prepare access the network?
SKILL.md names 2 domains. In commands or code: vivadicta.com; the agent is likely to contact it when it follows the instructions. As links in the text: icloud.developer.apple.com. This is read from the text; nothing was executed.
Is Release Prepare safe to install?
Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.
What licence does Release Prepare use?
Release Prepare is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.
How many tokens does Release Prepare use?
About 7.8k tokens (SKILL.md is roughly 31k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 898 tokens, read only when the agent opens those files.
What are the alternatives to Release Prepare?
Skills that share tags, products or a category with Release Prepare: Pypi Maintainer (agentculture/culture, 113 stars), Kmp Starter Features (DevAtrii/Kmp-Starter-Template, 167 stars), Android Readme Screenshot Studio (permissionlesstech/bitchat-android, 7.8k stars) and Release Asc CLI (rorkai/App-Store-Connect-CLI, 7.7k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.
Who maintains Release Prepare?
n0an (a GitHub user) maintains it in n0an/VivaDicta, which has 131 GitHub stars. The repository holds 20 skills in this directory. The repository was last updated on October 5, 2026.
Source: n0an/VivaDicta on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.