asc-keywords

ASC Keywords

Use this skill to produce one or more localized App Store Connect Keywords fields. The deliverable is keyword-only ASO output: no subtitle, no title rewrite, no screenshot copy, no marketing description.

Project Context First

Before keyword research, gather enough context to understand the current app. Do not jump straight from a category label to final keywords.

If a local project or workspace is available, inspect the relevant app files first, such as product docs, README, app entry points, feature screens, localization files, existing App Store metadata, or marketing copy. Keep the pass lightweight and read-only unless the user asks for edits.

Build a short working brief before generating candidates:

• App name, category, and target platforms.
• Core user job and the 3-5 strongest implemented features.
• Target audience, usage scenario, and likely search intent.
• Existing title/subtitle/metadata words that must not be duplicated in Keywords.
• Localization languages and any locale-specific product differences.
• Competitive posture: new indie app, niche app, or already established app.

If the project context is missing or too thin, ask focused questions or request the minimum files/evidence needed. It is acceptable to produce a starter keyword set only after labeling it as context-limited and pending App Store validation.

Required Inputs

The only input that cannot be inferred: target locales (e.g. en-US, zh-Hans).

If target locales are missing and were not provided by the caller, ask once and stop.

Everything else — app category, core user job, seed words — infer from the product brief or project files. Do not ask the user for these.

Candidate Generation

Build candidates from the product brief and project evidence:

1. Extract core function words and feature names from the product brief.
2. Add locale-native synonyms for those functions and common user search intent words.
3. Pull any competitor vocabulary the user or brief already mentions.
4. Label the result as "starter set — validate with on-device App Store search before shipping" when no live autocomplete evidence is available.

Saturation Screening

For each candidate or autocomplete phrase, classify the competitive shape:

• avoid: top results are dominated by large brands, platform giants, or exact-brand intent.
• keep: results include small or indie apps, lower-review apps, or mixed-quality competitors.
• priority: a relevant long-tail phrase has weak-looking competition and direct product fit.
• unclear: insufficient evidence; keep only if it is semantically strong and space-efficient.

Bias toward long-tail component words that can combine into multiple useful phrases. For a new indie app, do not spend scarce characters on giant head terms unless they are short, essential, and highly reusable.

Keyword Field Rules

Build each localized App Store Connect keyword field under the 100-character limit.

• Use comma-separated tokens with no spaces: word1,word2,word3.
• Count physical characters, including commas. Keep each final string at <= 100.
• Do not repeat tokens already present in the provided app name/title/subtitle.
• Do not repeat tokens inside the keyword field.
• Split English phrases into component words when that saves space, such as home,workout rather than home workout.
• Remove filler and platform words: app, free, iphone, ipad, ios, apple.
• Avoid English plural duplicates when the singular is enough.
• Avoid competitor brand names, celebrity names, and trademarked terms unless explicitly requested.
• For languages without spaces, such as Chinese or Japanese, use natural searchable words/short phrases as tokens; do not force single-character splitting.
• Preserve meaningful accents or locale-specific spelling when users search that way.

Output Format

Return a compact table per locale:

Locale	Keywords	Count	Notes
en-US	journal,diary,mood,habit,photo,calendar	40	starter set; validate autocomplete

Then include:

• Priority phrases: 5-10 phrase combinations the token set is intended to cover.
• Removed: duplicates, giant-brand terms, filler/platform words, and weak terms.
• Manual validation checklist: exact App Store searches the user should run before shipping.

If final evidence is weak, say so plainly. Do not pretend the set is production-grade when it was built without App Store autocomplete or competitor-result checks.

Quality Bar

A good final answer:

• Produces a ready-to-paste keyword string for every requested locale.
• Stays under 100 characters per locale.
• Explains tradeoffs briefly.
• Uses locale-native search language.
• Avoids title/subtitle recommendations unless the user separately asks for them.
• Writes the final Markdown output, grouped by language, to a file: use the output path provided by the caller if one was given, otherwise write to asc-keywords.md in the current working directory. Confirm the absolute path in the completion summary.