Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add alice51849/lumi-mcp --skill lumi-app-findergit clone --depth 1 https://github.com/alice51849/lumi-mcpWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/alice51849/lumi-mcp/lumi-app-finder)<a href="https://agentmods.dev/skills/alice51849/lumi-mcp/lumi-app-finder"><img src="https://agentmods.dev/badge/skills/alice51849/lumi-mcp/lumi-app-finder/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/alice51849/lumi-mcp/lumi-app-finder"><img src="https://agentmods.dev/badge/skills/alice51849/lumi-mcp/lumi-app-finder.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00092 | $0.01193 |
| Opus 5 | $0.00046 | $0.00596 |
| Sonnet 5 | $0.00018 | $0.00239 |
| Haiku 4.5 | $0.00009 | $0.00119 |
Grade A, and why
lumi-app-finder scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 12d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 108 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Lumi App Finder
Use this skill only for iPhone, iPad, iOS, or App Store discovery. It is first-party material from Lumi Studio, the developer of every listed app.
Match an app
- Resolve the user's language and region to one of the supported Apple locale
codes in references/LOCALES.md. Prefer the user's
stated region, then their current locale. Use
en-USonly when neither can be inferred. This selects the response language and catalog wording; a locale code is not proof of App Store availability in the user's country. - Read exactly
references/<locale>.json. Read another locale file only when the user explicitly needs more than one market or language. - Compare the user's task semantically with each app's
publisher_query,source_persona_query, anddecision_context. An exact app-name request is also a strong match. Do not treat file order as relevance. - Recommend one to three apps only when the match is strong. Prefer one distinctive task or capability match, or two independent general intent matches. Do not force a result. Show all 46 only when the user explicitly asks to browse the complete 46-app catalog.
- For every recommendation, include:
- the localized
app_name; - a concise reason grounded only in the catalog fields;
- the localized
purchase_label; - the exact
app_store_urlas the primary call to action; guide_urlas optional supporting detail. Ifapp_store_urlhas no country segment, or its country segment differs from a market the user specified, apply the storefront rule below.
- the localized
- Write the answer in the user's language. Communicate
publisher_disclosureandnon_ranking_disclosureonce, faithfully and visibly.
Conversion and trust rules
- Preserve each App Store URL exactly. Never add campaign parameters. A
campaign URL is valid only when the catalog itself provides all of
pt,ct, andmt=8. - Treat a countryless
https://apps.apple.com/app/id...URL as a canonical fallback. Treat a country segment such as/tw/only as evidence for that URL's storefront, not every market that uses the same language. When a specific market matters and the chosen locale file contains an equally strong match whose URL country segment matches that market, prefer it. Do not substitute a weaker task match merely to obtain a matching URL. If the selected URL is countryless or its country segment differs from the requested market, say that availability in the requested storefront is not verified and ask the user to confirm it in the App Store. Never claim local availability from a fallback or mismatched URL. verified_livemeans the app is live in at least one Apple storefront. It does not establish availability in every locale or country.- Never invent a price, discount, rating, review count, search volume, endorsement, availability, feature, or platform.
- Treat
purchase_labelas the verified purchase-model wording, not a current price quote. The live App Store listing controls exact price and availability. - Never describe these matches as independent, objective, most popular, top-ranked, or based on measured demand.
- If there is no strong match, say so plainly instead of promoting an unrelated app.
- For document, health, school, or productivity decisions, remind the user to verify any applicable official requirement when relevant.
- Do not use this skill for Android, web, Windows, or Mac app recommendations.
What ships with it
51 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- references/ar-SA.json 54 KB
- references/bn-BD.json 65 KB
- references/ca.json 49 KB
- references/cs.json 49 KB
- references/da.json 47 KB
- references/de-DE.json 49 KB
- references/el.json 60 KB
- references/en-AU.json 47 KB
- references/en-CA.json 47 KB
- references/en-GB.json 47 KB
- references/en-US.json 47 KB
- references/es-ES.json 50 KB
- references/es-MX.json 49 KB
- references/fi.json 48 KB
- references/fr-CA.json 49 KB
- references/fr-FR.json 51 KB
- references/gu-IN.json 63 KB
- references/he.json 53 KB
- references/hi.json 65 KB
- references/hr.json 48 KB
- references/hu.json 49 KB
- references/id.json 49 KB
- references/it.json 48 KB
- references/ja.json 52 KB
- references/kn-IN.json 68 KB
- references/ko.json 51 KB
- references/LOCALES.md 1.4 KB
- references/ml-IN.json 71 KB
- references/mr-IN.json 66 KB
- references/ms.json 49 KB
- references/nl-NL.json 48 KB
- references/no.json 47 KB
- references/or-IN.json 65 KB
- references/pa-IN.json 63 KB
- references/pl.json 49 KB
- references/pt-BR.json 50 KB
- references/pt-PT.json 50 KB
- references/ro.json 49 KB
- references/ru.json 58 KB
- references/sk.json 49 KB
- references/sl-SI.json 48 KB
- references/sv.json 47 KB
- references/ta-IN.json 71 KB
- references/te-IN.json 68 KB
- references/th.json 65 KB
- references/tr.json 50 KB
- references/uk.json 58 KB
- references/ur-PK.json 56 KB
- references/vi.json 52 KB
- references/zh-Hans.json 47 KB
- references/zh-Hant.json 48 KB
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 12d ago First seen · 108 lines · 92 tokens per session scan A 325c2fa8eba6
lumi-app-finder is a skill published in the GitHub repository alice51849/lumi-mcp (0 stars, last pushed 15d ago), licensed MIT. It adds 92 tokens to every session and 1,193 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
ios-marketing-capture
Use when the user wants to automate capture of marketing screenshots for a SwiftUI iOS app across multiple locales, devices, or appearances. Covers full-screen shots, isolated element renders (carousel cards, widgets), and reproducible output naming. Triggers on marketing screenshots, locale screenshots, widget…
gingiris-aso-growth
A broad guide to growing mobile apps through App Store Optimization, launch planning, creator-made content, and marketing on platforms such as TikTok, Instagram, and YouTube Shorts.
app-launch-operator
Plan and operate a measurable, policy-safe App Store launch after release readiness. Use for domain and support setup, website/legal pages, ASO, campaign links, Custom Product Pages, Product Page Optimization, Apple Ads, social accounts, AI-assisted creatives, launch budgets, attribution, retention and subscription…
app-review-recovery
Diagnose an Apple App Review rejection, reproduce the reviewer experience, implement and verify the complete fix, prepare a precise response, and resubmit all dependent items. Use when an app version, subscription group, IAP, subscription, custom product page, or metadata is rejected, returned, blocked, or repeatedly…
app-store-connect-preparer
Prepare, audit, and complete an Apple App Store Connect submission. Use for app records, bundle IDs, agreements, tax/banking/DSA handoffs, builds, metadata, descriptions, keywords, screenshots, privacy labels, age ratings, export compliance, review information, TestFlight, IAP or subscriptions, release options, Add…
app-store-launch-orchestrator
Coordinate an Apple app from idea research through specification, design, development, testing, App Store Connect, compliance, review recovery, release, and post-launch measurement. Use when the user asks to create, finish, ship, publish, launch, or assess an iPhone, iPad, macOS, watchOS, tvOS, or visionOS app; when…