--- name: store-listing-optimization description: "Design or audit one channel listing and creative-conversion artifact for App Store, Google Play, Steam, console/PC stores, Microsoft Store, browser extensions, or direct download: positioning, title/subtitle/description, screenshots, capsule/icon/key art, trailer, proof, pricing/disclosures, privacy/compatibility claims, localization, accessibility, experiments, review/reputation context, and exact asset QA. Use for listing conversion; use Distribution Readiness for submission/certification and Marketing Automation for the full channel portfolio." --- # Store Listing Optimization Produce a **Channel Listing and Creative Sequence** that communicates truthful value before install, wishlist, trial, or purchase and can be generated, localized, tested, and updated across exact channel variants. ## Atomic boundary Own one channel listing's positioning-to-asset narrative, metadata, creative sequence, proof/disclosures, localization, accessibility, asset QA, experiment, and conversion/trust measurement. Do not own store submission/certification, whole-product positioning, public review solicitation, paid acquisition, or the multi-channel creative/spend control plane. For every artifact, record `artifactVersion`, `artifactRevision`, and `artifactState`; never put `artifactDigest` on the top-level artifact. Use the [shared artifact envelope](references/product-artifact-envelope.schema.json) for structured drafts and sealed artifacts. Consume product thesis, canonical product/build truth, a planning Product Program Manifest revision when present, and only the Marketing, pricing/payment, review/reputation, analytics, Product Asset, or distribution facts applicable to that exact revision. The initial request revision must not consume its future Product Asset pack; a later final listing revision may. A sealed input additionally requires `artifactDigest` and `digestRule: sha256-exact-bytes`; every typed input requires `fulfillsHandoffId`, while a draft contains no digest fields. ## Agent-first invariant Build the complete listing narrative, selection/order, metadata, channel templates, asset requests, experiment variants, validation, versioning, rollback, and readback now. Marketing owns campaign briefs/concepts; `product-asset-production` owns deterministic capture, localized/accessibility variants, exact rendered files, rights/provenance, and file QA. This skill emits an asset request in an earlier revision and consumes the returned pack only in a later exact revision; it must not create a second production source of truth. Construction is separate from publication. A dormant channel generates no public page and claims no compatibility until authority and the exact pack pass. ## Workflow 1. Define exact product/build/content identity, channel/territories/locales, primary audience and job, positioning promise, conversion event, price/model, device/input/age modes, current listing evidence, and non-goals. 2. Read `references/store-listing-optimization-patterns.md`. Retrieve the channel's current metadata, asset, content-rating, pricing/disclosure, experiment, and localization contract. 3. Map audience objections and proof to a narrative sequence: distinctive outcome/mechanism, first-value/core interaction, depth/progression or workflow, social/collaboration where real, content/utility, trust/quality, business model/compatibility, and call to action. 4. Specify title/subtitle/short/long text, keywords/tags/categories, icon/key/ capsule art, selection/order and stable production requests for screenshots by device/input, video/trailer, captions/transcript, alt text, pricing/IAP/subscription/ad disclosures, privacy/compatibility, and support links. Every claim maps to exact shipped behavior. 5. Build localization briefs per locale: intent, search language, cultural meaning, reading direction, typography, image/text constraints, age modes, price/unit/date, and native QA. Do not reuse one translated screenshot set. 6. When exact media is selected, consume the Product Asset Production Pack in a new listing revision; validate dimensions, safe zones, compression/color/ audio, device chrome, legibility, motion/captions, no unreleased UI, rights, claim evidence, channel rules, and exact package digests. 7. Define randomized/native experiment where supported, immutable variant and traffic identity, exposure/contamination, install/wishlist/purchase plus retention/refund/review/support countermetrics, confidence, and stop rules. 8. When publication is selected, hand the exact final listing revision and accepted asset pack when applicable to `software-distribution-readiness`; verify submitted/live metadata and assets, then archive/supersede the prior version. ## Source verification Retrieve current channel metadata/asset/experiment, ratings/reviews, price/IAP/ subscription/ad disclosures, privacy/data-safety, age/content rating, compatibility, accessibility, localization, IP/rights, and regional rules. Unknown/stale channel authority blocks submission of that adapter. ## When not to use - Use `software-distribution-readiness` for eligibility, signing, submission, reviewer/certification, staged rollout, withdrawal, and live channel evidence. - Use `marketing-automation-blueprint` for the whole SEO/creator/community/ lifecycle/paid creative portfolio, spend, attribution, and shutdown. - Use `product-asset-production` to create deterministic captures, rendered screenshots/art/video, localized/accessibility variants, and the exact media pack requested by this listing. - Use `market-research-synthesis` when product positioning and audience demand are not yet evidence-bounded across the market. - Use `review-solicitation-policy` for public rating/review request eligibility and `product-feedback-learning-loop` for private feedback, authorized review ingestion/responses, support routing, and product close-loop. ## Guardrails - Never show unreleased/nonexistent UI, fake rankings/reviews/awards, hidden prices/renewal/ads, misleading gameplay, unsupported devices, or privacy/ accessibility claims without exact evidence. - Search optimization cannot reduce human clarity or stuff irrelevant terms, competitor marks, or deceptive claims. - Do not sentiment-gate, incentivize, fabricate, filter, or steer ratings; review context is observed evidence, not listing manipulation authority. - Separate creative conversion from product quality. A higher install rate with worse activation, retention, refunds, support, or trust is not a win. - Autonomous generators cannot publish, change price/claims/rights, or promote a variant without exact validation and authority. - Do not capture, render, localize, or hand-edit product media; request a new canonical `product-asset-production` pack revision. ## Output contract Return one typed Channel Listing and Creative Sequence with: 1. draft identity or sealed envelope, exact product/channel identities, audience/job, positioning, conversion, evidence labels, assumptions, and non-goals; 2. objection/proof/narrative sequence and metadata field contract; 3. icon/key/capsule, screenshot/device/input, trailer/video, caption/transcript/ alt-text, disclosure, compatibility, support selection/order, and stable Product Asset Production request plus accepted pack references; 4. per-locale search/meaning/culture/RTL/typography/image/price/date/age brief; 5. channel-authority, rights, claim, dimensions/safe-zone, media, accessibility, and exact-package QA evidence; 6. experiment variants, assignment, metrics/countermetrics, confidence, stop, rollback, and live readback; 7. digest-pinned distribution/marketing/review/analytics handoffs. Complete only when every live claim/asset is exact-build truthful, localized and accessible, the package is reproducible by digest, and conversion gains survive downstream quality/trust countermetrics.