Aso Audit

ASO Audit

Reuse Growth Context

At the start, read .agents/growth-context.md when it exists. Reuse only product, customer, market, outcome, constraint, evidence, and routing fields whose scope, definition, source, and date remain compatible; state what is reused and surface conflicts or staleness before asking for decision-changing gaps. The file grants no system access or execution authority, and this Skill must not silently rewrite the primary diagnosis.

Audit store listings, not product implementation, release submission, or app code. Route those requests to engineering or release operations; state that a later audit needs a dated listing capture, store identifier or canonical URL, country, locale, and requested surfaces.

Compose The Smallest Scope

Choose one primary mode from the decision requested, then add only necessary modifiers:

• listing primary: assess a bounded Apple App Store or Google Play product page when the decision concerns the listing overall.
• focused primary: decide a question about only named listing surfaces, such as the first three screenshots; do not expand to metadata, reviews, or the app.
• performance primary: investigate a reported store-performance signal. Treat it as a signal until dated, segmented measurement verifies it; keep performance primary when adding comparison or surface bounds.
• competitive modifier: compare supplied listings using the same store, fixed card, country, locale, and date. Label the work competitive, but retain the decision's primary mode; a performance comparison is performance + competitive, not competitive primary. Listing-page evidence cannot establish query rank.
• focused bound: constrain a performance or competitive assessment to the named surfaces without changing its primary decision. For example, a performance diagnosis comparing only two screenshots is performance + competitive, focused to those screenshots.
• out-of-scope: own implementation-only requests for onboarding, deep links, release submission, or unrelated engineering. Give a concise ownership handoff instead of an audit. For mixed requests, audit only the ASO portion and hand off implementation work.

Establish Scope And Evidence

1. Identify the store from a canonical URL or stable identifier: Apple ID/apps.apple.com routes to Apple; package/play.google.com routes to Google Play. Record app, canonical ID/URL, country or storefront, locale, capture/retrieval date, mode, requested surfaces, and limits. Do not substitute another store, market, locale, or historical capture.
2. Start public-first with supplied dated captures or bounded, read-only public pages. Do not require App Store Connect, Play Console, attribution, keywords, acquisition, ranking, conversion, or experiment access before beginning. Obtain authorization before private-console access or high-volume collection; keep access read-only and within the agreed scope.
3. Inventory observable surfaces and failed renders. Apple: name, subtitle, promotional text when displayed, description, screenshots, preview when displayed, ratings, reviews, age/content, release notes, and localization. Google Play: title, short description, full description, category, feature graphic when displayed, screenshots, promotional video when displayed, ratings, reviews, content rating, Data safety, release information, and localization.
4. Label every statement: verified for a dated observation or attributable export; inferred for a reasoned explanation or expected effect; unavailable for applicable missing/inaccessible/unrendered evidence; not applicable for another platform or excluded scope. A loading skeleton, HTTP error, or partial page proves only that rendered state.
5. Use optional owner evidence to refine an already-started audit: dated listing history, query capture, keyword export, acquisition/source/territory time series, attribution, experiment plan/results, and review samples. Ask for the smallest evidence that would change a conclusion.

Assess With Fixed Cards

1. Read [Apple card](references/apple-audit.md) or [Google Play card](references/google-play-audit.md), then apply only its fixed atomic checks and six shared dimensions: Discovery, Metadata, Creative, Trust, Localization, and Experimentation. Use platform terms; Apple subtitle is not Google Play short description, and cross-platform fields are not applicable.
2. Use [evidence and scoring](references/evidence-and-scoring.md): assessed checks receive exactly pass 100%, partial 50%, or fail 0% of the fixed maximum. Never reweight for brand fame, asset quantity, aesthetics, executive preference, or absent private data.
3. Keep unavailable applicable checks in coverage but out of score math; they earn and lose zero points. Exclude not-applicable checks from both denominators. Report score = assessed earned / assessed max * 100 and coverage = assessed applicable max / all applicable max * 100, each to two decimals; show Not scored when its denominator is zero.
4. Define the full membership of each collection-level check before scoring. If membership is unknown or any applicable member is unavailable, the parent is unavailable; record member observations but do not subscore or partially score the collection.
5. Produce the required trace: check ID | state | result | earned/max | finding ID. Give every partial/fail one stable, evidence-linked finding; list unavailable checks in a separate register with the smallest next collection step. No missing-data penalty.

Diagnose And Recommend

1. Inspect public copy, creative, visible trust signals, locale fit, and dated listing changes. For screenshots, record supplied or observed viewport, rendered size, message size, contrast, obstruction, first-use-case clarity, and narrative sequence. Label an expected conversion effect inferred; do not claim conversion, rank, installs, causality, or uplift without dated measurement.
2. For performance work, separate verified measurement (metric, segment, market, window) from inferred cause. Prioritize a validation sequence: before/after listing history, acquisition source and territory series, query/visibility evidence, then controlled experiment result.
3. Order findings by verified fail, verified partial, inferred adverse partial; then fixed-point deduction, validation dependencies, and lower effort. Keep public-surface changes separate from evidence-collection tasks. Each finding has ID, scoped source/date/check, observed impact, action, observable completion proof, and S/M/L effort.
4. Treat popularity and premium appearance as context, never proof of discoverability, relevance, quality, conversion, or installs. Treat platform requirements as verified only when a current primary source supplies its URL and retrieval date; add publication/update date only if exposed. Otherwise mark the exact rule unavailable and point to [platform sources](references/source-notes.md).

Handle Market Transfer And China

Treat each store, country or storefront, locale, listing version, and capture date as a new evidence boundary. A listing, score, query capture, experiment, conversion signal, or recommendation from one boundary does not validate another. Do not substitute another store, market, locale, or historical capture when the requested surface is unavailable.

For China, keep market, language, locale, Apple App Store or Google Play product, app availability, app distribution, listing, account, submission, payment, data, content, and locally accountable review separate. When Simplified Chinese is requested, write native Simplified Chinese. Do not infer platform or app availability, submission eligibility, compliance, ranking, conversion, installs, or performance. Use current direct official sources for platform-specific claims or mark them unavailable; do not provide legal or regulatory conclusions.

An audit authorizes local artifacts only. Do not access App Store Connect, Play Console, analytics, attribution, experiment, publishing, or release systems; change listing copy or assets; submit; publish; or deploy without separate task-level authorization and capability review.

Deliver

Present scope and limits; evidence states; fixed-card point trace; score and coverage with numerators/denominators; unavailable register; findings/actions/completion; prioritized roadmap; and validation plan.

For every in-scope listing, focused, or performance audit, generate one standalone offline HTML report in the same task as the default primary artifact, whether or not the requester explicitly asks for HTML. Apply the same rule when competitive modifies an in-scope audit. Do not ask whether to generate the report, defer it to a later step, or claim completion before the file exists and has been checked. Follow [HTML report](references/html-report.md), write {app-slug}-{store}-aso-audit.html, and return its local path with a concise conversational summary. Missing private evidence, rejected visual assets, partial page rendering, or a low coverage score does not block the file: preserve the limits, unavailable register, and inert placeholders inside the report.

Skip HTML only when the requester explicitly opts out, asks for a non-HTML-only result, or the request is entirely out-of-scope. If the environment cannot write a local file, state that blocker in the current response and provide the complete report content without promising a later artifact. Hand implementation-ready listing-copy or creative briefs to the owning marketing/content/design team; hand app-code and release work to engineering/release operations.

Keep One Output Language

Use the requested output language consistently across headings, prose, tables, labels, and actions. When no language is explicit, match the user's dominant language; market, locale, platform, and source language do not override it.

For Simplified Chinese, write natural Simplified Chinese and translate ordinary business or analytical jargon instead of embedding English words such as owner, brief, listing, cohort, baseline, benchmark, guardrail, gate, finding, roadmap, workflow, and handoff. Keep only proper names, standard acronyms after a Chinese first-use definition, machine tokens or IDs, code, formulas, filenames, URLs, and exact quotations where necessary.

For English, use idiomatic English and do not add Chinese glosses except for proper nouns or quoted source text. Use multiple languages only when explicitly requested, and keep each version in a separate labeled section rather than mixing languages within sentences or tables. Do not alternate languages for emphasis or perceived expertise.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: krillinai
• Source: krillinai/growth-skills
• License: MIT
• Homepage: www.clawee.ai/

Install and usage instructions live in the source repository linked above.