--- name: aso-audit description: Runs an App Store Optimization (ASO) audit of an iOS or Google Play app and returns a per-storefront report with PASS / FIX / UNKNOWN for every check. Uses measured data from the Storelift MCP server (keyword ranks, rivals above you, rank history, store page, reviews, charts, AI visibility) when it is connected, and falls back to a checklist over the public store listing when it is not. Use when the user asks for an "ASO audit", "audit my App Store listing", "check my app's ASO", "why is my app not ranking", "ASO checklist" or "ASO report". --- # ASO audit This skill audits one app's store presence and returns a report the user can act on. It has two modes: - **Mode A — measured.** The Storelift MCP server is connected. Ranks, rivals, history, reviews and charts come from Storelift's nightly measurement of the public App Store and Google Play. - **Mode B — checklist.** No MCP server. The audit checks the public listing against the stores' field rules and the user's own text. It reports no ranks and no search demand, because none were measured. ## Ground rules (read first, apply everywhere) 1. **Never invent a number.** No search volume, popularity, difficulty, download estimate or "ASO score" that did not come back from a tool or from the user. If a check needs data you do not have, mark it UNKNOWN and say what would answer it. 2. **Keep the three keyword states apart.** In `get_keywords`: - `rank: ` — measured, the app was at that position - `rank: null` with `measured: true` — measured, the app was **not in the results** - `measured: false` — the search **failed**; nothing is known Report them as three separate counts. Never write an unmeasured term as "rank 0" or "not ranking". 3. **One storefront at a time.** A rank belongs to one country and one platform. Do not average across countries or merge iOS with Google Play. 4. **No score.** Report counts of checks (for example "14 PASS, 5 FIX, 3 UNKNOWN"), not a weighted score. A single number would hide which part failed. 5. **Read-only.** This skill does not change the listing, the keyword field or the tracking setup. It writes a report; the user decides what to ship. ## Step 0 — Pick the mode Call the Storelift tool `list_apps` (in Claude Code it appears as `mcp__storelift__list_apps`; other clients may prefix it differently). - It returns apps → **Mode A**. Ask which app and which storefronts to audit if the user did not say; default to the app's first country and iOS. - The tool does not exist, or it returns an error about the API key or the plan → **Mode B**. Tell the user once, in one sentence, that the measured part needs the Storelift MCP server (Pro or Studio plan), and continue with the checklist. Setup: https://storelift.net/guide/storelift-mcp-server/ Ask the user for anything the store does not publish and you will need: the current **iOS keyword field** (Apple does not publish it; no tool can read it) and, in Mode B, the app's App Store or Google Play URL. ## Mode A — measured audit with the Storelift MCP server The server has eight read-only tools. Every tool except `list_apps` takes `appId` (from `list_apps`) and `country` (two-letter code such as `us`, `tr`). `get_keywords` and `get_rivals` also take `platform` (`ios` or `android`, default `ios`). Run these per storefront, in this order: ### A1. Keyword coverage — `get_keywords` Returns every tracked term with `measured`, `rank`, `demand` (iOS only; null when not measured) and `searchDepth`. Report: - counts: ranked 1–10, ranked 11–50, ranked below 50, measured but absent, could not be measured - the terms with the highest `demand` where the app is absent or below 50 — these are the gaps worth the most attention - if `live` is false, the storefront has no measurement yet; say so and stop this step for that storefront Do not print a demand column for Google Play; demand is an App Store signal and an empty column would read as "zero demand". ### A2. Who is above you — `get_rivals` Returns the apps ranking above this app, each with the terms it beats you on, `theirRank`, `ourRank` and its rating count. Report the three rivals that sit above you on the most terms, and for each the terms and both ranks. `ourRank: null` means the app does not appear for that term at all. If a term is already in the app's own name or subtitle and the app still ranks far below rivals with many more ratings, say that more keywords will not fix it; ratings and installs are the gap. ### A3. What moved and why — `get_history` + `get_store_page` `get_history` returns `[day, rank]` series per term (the window depends on the plan). A null point is a day the app was absent, not a bad rank. `get_store_page` returns a `timeline` of dated changes detected on the listing: new version, changed screenshots, new in-app events, new editorial placements. Line up large rank moves with timeline entries on or just before the same day and report the matches. Where nothing lines up, say the cause is not visible in this data instead of guessing. ### A4. The listing itself — `get_store_page` For iOS it returns ratings (average, total, 1–5 star histogram), the current version and its release notes, the screenshot set counted per device shelf (iPhone and iPad), in-app events, editorial placements and the similar-apps shelf. For Google Play it returns the Play page (install count, rating histogram, ad and in-app purchase flags). `measured: false` means the page was never read; it is not an empty page. Check and report: - **Screenshots:** how many per shelf, out of the 10 Apple allows per shelf. Fewer than 10 is a FIX only if the user has more to show; the count itself is a fact, not a verdict. - **Version recency:** date and content of the last release notes. - **Ratings:** average, total, and whether the histogram is weighted to 1-star. - **In-app events:** any running or scheduled. None is a note, not a failure. - **Similar-apps shelf:** the categories of the apps Apple shows next to this one tell you which shelf Apple has put the app on. If they are off-topic, flag the category choice for review. The tool does **not** return the name, subtitle or keyword field. Read name and subtitle from the public listing (`https://apps.apple.com//app/id`, where `iosId` comes from `list_apps`) if you can fetch web pages, or ask the user to paste them. Then run the metadata checks from Mode B (B1) on them, using the measured terms from A1 as the term list. ### A5. Reviews — `get_reviews` Returns recent App Store reviews (`list`), the Google Play reviews (`android`) and `newCount`, the number that arrived since the previous measurement. `newCount: null` is an older record, not zero; `android: null` means Play was never measured. Report recurring themes with a count and one short quote each: crashes and bugs, pricing and paywall complaints, ads, missing features. Flag 1–2 star reviews that name a bug that the latest release notes do not mention. ### A6. Charts — `get_charts` (App Store only) Returns current positions per chart list (free, paid, grossing; overall and in category) with history. Google Play publishes no chart list, so there is no Play row; do not mark that as a failure. ### A7. AI visibility — `get_ai_visibility` Returns whether Claude, ChatGPT and Gemini name the app when asked about its category, per engine. The data is filled on the Studio plan; on other plans report the step as UNKNOWN (not measured). Treat it as an observation, not a ranking: model answers vary between runs, so read the trend. ## Mode B — checklist over the public listing Use the listing the user points to and the text they paste. Report every line as PASS, FIX (with the exact text that fails) or UNKNOWN. ### B1. App Store metadata | Check | Rule | |---|---| | Name | ≤ 30 characters | | Subtitle | ≤ 30 characters; describes what the app does, not a slogan | | Keyword field | ≤ 100 characters; comma-separated; no space after commas | | No repeats | No word appears in more than one of name, subtitle and keyword field; the store indexes the three together, so a repeat spends characters for nothing | | Word coverage | For each term the user wants to rank for, every word of the term appears somewhere across name + subtitle + keyword field (the store matches words across the three fields, not only exact phrases) | | Apple product names | No iPhone, iPad, Apple Watch or AirPods in name or subtitle (guideline 5.2.5); allowed in the keyword field | | Pricing language | No prices, discounts or "free trial" sale copy in any text field | | Other apps' names | No competitor app names or brands in any field | | Promotional text | ≤ 170 characters; not used for search, but editable without a new version | | Description | ≤ 4,000 characters; written for the reader who already found the app | Name, subtitle, keyword field and description change only with a new app version; promotional text can change on a live listing. ### B2. Google Play metadata | Check | Rule | |---|---| | Title | ≤ 30 characters; no emoji, no promotional or price wording, no "#1" or ranking claims | | Short description | ≤ 80 characters | | Full description | ≤ 4,000 characters; Play has no hidden keyword field, so the terms that matter must appear in the visible text, written as sentences | Do not apply Apple guideline numbers to a Play listing. ### B3. Screenshots and preview - Count per device shelf (iPhone, iPad; phone and tablet on Play). - The first screenshots are the ones seen in search results: check that the first one states what the app does in words a new user understands. - Captions are readable at thumbnail size and localized where the listing is. - Screenshots show the real app, current version. ### B4. Ratings and reviews - Average rating and total count per storefront (both are on the public page). - The app asks for ratings through the system prompt (`SKStoreReviewController` / `requestReview` on iOS, the In-App Review API on Android), not a custom "Do you like the app?" gate before it. On iOS the system prompt shows at most three times in 365 days, so ask at a moment of success. - Recent 1–2 star reviews: are there unanswered ones naming a bug? ### B5. Localization per storefront - List the storefronts the app is sold in and the localizations it ships. - In some App Store storefronts the store also indexes a second localization's metadata (for example en-GB in Türkiye). Where that second localization is empty, a second set of name/subtitle/keyword fields is left unused. Check the verified mapping per storefront at https://storelift.net/tools/storefront-languages - Keywords per locale are researched in that language, not translated from English. - Fixed strings stay untouched in every locale: product names, URLs, email addresses, numbers with units. ### B6. Checks that need measurement Without the MCP server, mark these UNKNOWN and name the free tool that answers one term at a time, with no account: - Current rank for one term → https://storelift.net/tools/rank-check - Keyword difficulty for one App Store term → https://storelift.net/tools/keyword-difficulty - Packing the 100-character keyword field → https://storelift.net/tools/keyword-field - Field lengths → https://storelift.net/tools/character-counter ## Report format Write one report per storefront (country × platform): ``` # ASO audit — — — Mode: A (Storelift MCP, measured ) | B (checklist, no measurement) Summary: PASS · FIX · UNKNOWN ## Fix first 1. 2. ... ## Keywords (Mode A) ranked 1–10: n · 11–50: n · 51+: n · absent: n · not measured: n | term | rank | demand | above you | ## Rivals above you (Mode A) ## Listing ## Reviews ## Charts (App Store only) ## AI visibility ## Not checked ``` Order "Fix first" by evidence, not by opinion: a high-demand term where the app is absent comes before a stylistic subtitle change. Every recommendation cites the data point or the rule it rests on. ## What this skill cannot do - Read App Store Connect or Google Play Console data (impressions, conversion rate, downloads, revenue). The Storelift MCP server reads the public stores, not your developer account. - Read the iOS keyword field of any app. Apple does not publish it. - Read Apple Search Ads data. The Storelift MCP server has no Apple Ads tools. - Report ranks for terms that are not tracked in Storelift, or history from before a term was added. Tracking starts when the term is added; there is no backfill. - Change anything in the store or in the Storelift account.