Appllama Mobile Design Skills
Research winning app patterns and build native-feeling mobile screens that survive simulator review.
Per-dimension scores and reasoning
The source identifies the MCP endpoint, account approval, per-call credit use, optional connection path, and code/trademark attribution, giving some transparency and provenance. The standalone design skill also avoids making external access universally mandatory. Deductions apply because the transmitted fields, retention practices, sensitive-data rules, and permission scope are not documented. A no-prompt installation mode is shown, while build or asset-generation steps lack confirmation checkpoints, a complete effects inventory, or rollback guidance. The use of npx skills@latest is not accompanied by pinning, integrity checks, or supply-chain controls.
The README is internally coherent about the roles of the two skills, how they combine, and when MCP is required. The MCP-dependent skill nevertheless relies on a paid account and an external service without compatibility versions, availability commitments, or documented degradation behavior. No failure messages, troubleshooting procedures, or recovery instructions are supplied. The simulator-verified bar is a stated workflow goal, not static evidence of execution reliability.
The intended audience and scenarios are unusually concrete: greenfield builds, screen improvement, flow research, checkout navigation, and animation review. Responsibilities and the design skill's standalone mode establish useful boundaries, but broad capability claims are not paired with a list of unsupported frameworks or tasks. Prompt examples give reasonable triggering cues, although actual trigger metadata is not included. Multiple agents and project/user-level installs are covered, with an Expo/React Native focus, but there is no OS, runtime, CLI, Expo, or React Native compatibility matrix.
The README has a clear skills table and sections for installation, connection, examples, and licensing. Installation documentation thoroughly covers default, agent-specific, global, single-skill, and manual paths. Names are consistent in the supplied files but lack a stability or migration policy. Examples are useful, though no FAQ is present. Paid MCP access, credit charging, and a standalone alternative are disclosed, but known limitations are not systematically documented. The complete MIT license and trademark qualification justify full license credit. No versioning scheme, changelog, or release history is shown. Appllama and Antmind Ventures are identified, but named maintainers, support routes, response expectations, and governance are absent; unknown publisher verification does not itself cause a deduction.
The material defines a usable research-to-design-to-simulator workflow, native-mobile output goals, and several practical prompts. Combining pattern research with implementation guidance plausibly adds value over generic design advice. Scores are reduced because no produced screens, sample project, or statically inspectable quality comparison is included. Cost disclosure covers the Pro requirement, one credit per call, free balance checks, and a non-MCP path, but omits pricing, expected calls per task, time requirements, and measured benefits.
Major quality claims—studying top-grossing apps, reaching an unfaultable simulator bar, and producing native-quality screens—are asserted in the README without a study list, methodology, example output, or corroborating implementation material. README and LICENSE cross-corroborate only the MIT license and legal attribution, not functional results. Marketing claims, intended process, and demonstrated outcomes are not consistently separated, so all three criteria receive only limited credit.
- MCP research sends requests to an external Appllama service and consumes paid credits; confirm transmitted content, retention practices, and account permissions before submitting screenshots, code, or app materials.
- The installation command uses npx skills@latest without pinning the CLI or skill revision. Review the resolved version and contents before adoption and pin a revision where practical.
- No actual outputs, tests, simulator recordings, or reproducible top-grossing-app research method are supplied. Treat native-quality and simulator-verified language as unverified positioning rather than independently established results.
- Rollback, failure handling, compatible versions, and a support-response path are undocumented. Establish review, backup, and recovery controls before use in consequential projects.
What does this agent do, and when should you use it?
This repository contains two complementary agent skills: `appllama-usage` and `appllama-app-design-skill`. The first uses the Appllama MCP to study screens, flows, UI patterns, and commercial context from top-grossing mobile apps, with playbooks for new products, screen improvements, and focused flow or element research. The second guides an agent in building native-feeling Expo and React Native screens, covering Apple HIG fidelity, semantic colors, native controls, navigation, motion, and generated image assets. Its stated delivery boundary is a working simulator flow recorded and inspected frame by frame, rather than a static mockup or screenshot alone. The research skill requires Appllama MCP access, while the design skill can operate independently and becomes more research-informed when MCP is connected.
appllama-usage calls the Appllama MCP to examine real screens, user flows, interface elements, and revenue and download context from successful mobile apps, then uses that evidence to identify a category's design language. It supplies workflows for building an app from scratch, improving an existing screen, and researching particular flows or elements. appllama-app-design-skill turns those findings into Expo and React Native screens while directing Apple HIG alignment, semantic colors, native controls, push-versus-replace navigation, sheets, overlays, and one-way states such as preventing a paid user from returning to a paywall. Its motion rules address whether an element should animate, whether springs preserve finger velocity, and whether animation work stays off the JS thread. The process finishes by running and recording complete flows in a simulator and scrubbing the result frame by frame.
- An Expo or React Native team starting a mobile product and wanting to study commercially successful apps in the same category before implementation.
- A designer or developer improving an existing screen from a screenshot, source code, or an Appllama Copy Screen ID reference.
- A product team investigating how leading fitness or habit apps structure onboarding, justify each step, and position a paywall.
- A developer wiring a checkout or subscription flow who needs explicit decisions about pushes, replacements, sheets, overlays, and post-payment back navigation.
- A mobile team reviewing motion for unnecessary effects, JS-thread execution, or springs that fail to carry gesture velocity.
- A developer who does not need Appllama research but wants a reusable native-quality and simulator-validation discipline for Expo or React Native.
What are this agent's strengths and limitations?
- Separates research and implementation into two composable skills that can also be installed independently.
- Grounds research in real screens, flows, UI patterns, and revenue and download context rather than generic design guidance alone.
- Addresses mobile-specific implementation details including Apple HIG fidelity, native controls, semantic colors, navigation semantics, and irreversible flow transitions.
- Uses full-flow simulator recordings and frame-by-frame motion inspection as its stated quality bar instead of relying only on screenshots.
- Explicitly documents use with Claude Code, Cursor, Codex, and the skills installer used by more than 70 agents.
appllama-usagedepends on the Appllama MCP, which requires a Pro plan and consumes credits for most calls.- The simulator-based acceptance loop requires a runnable mobile project and simulator, making adoption heavier than a static design-advice workflow.
- The documented build focus is Expo and React Native; the source provides no equivalent support claim for native Swift, Kotlin, or other cross-platform stacks.
- The supplied material contains no automated test results, benchmarks, or finished sample applications that quantify output quality.
- Credits reset monthly, so research-heavy workflows must account for call volume and available allowance.
How do you install or deploy this agent?
From the project root, run npx skills@latest add appllama/appllama-skills. For a non-interactive Claude Code and Cursor installation, run npx skills@latest add appllama/appllama-skills -a claude-code -a cursor -y; add -g for a user-wide installation. To install only the design skill, run npx skills@latest add appllama/appllama-skills --skill appllama-app-design-skill; substitute appllama-usage to install only the research skill. For a manual installation, clone https://github.com/appllama/appllama-skills and copy skills/* into the agent's project or user skill directory, such as ~/.claude/skills/. Research features also require adding https://mcp.appllama.io/mcp as a custom connector in an MCP client and approving it with an Appllama account; MCP access is included with Appllama Pro.
How do you use this agent?
After installation, describe the task in Claude Code, Cursor, Codex, or another compatible skills-based agent. A documented first prompt is: Build me a habit tracker. Study the top-grossing habit apps first and don't stop until every screen survives the simulator comparison. For an existing interface, attach a screenshot or code, or provide a Copy Screen ID from appllama.io. Connect the MCP before requesting Appllama-backed research; get_credits checks the remaining balance for free, while every other MCP call consumes one credit. If using only appllama-app-design-skill, MCP is optional, although carrying out its build and validation process still calls for an appropriate Expo or React Native project and simulator environment.
FAQ
Is an Appllama Pro subscription mandatory?
appllama-app-design-skill. It is required for the Appllama MCP used by appllama-usage.How are MCP calls charged?
get_credits, which is free. The full credit allowance resets on the first day of every month.Does it produce a runnable mobile app by itself?
Can the two skills be installed separately?
--skill appllama-app-design-skill for the design skill or --skill appllama-usage for the research skill.