benjitaylor/skills

What it is

Skills is a collection of skill packs for AI coding agents, focused on design and motion work. The first published skill, Stageshow, helps an agent produce faux product UI that resembles a real product: stills, feature cards, cursor demos, and bento grids, all built in React. The pack bundles composition and motion guidance, a kit to copy from, examples, and tools that render the result so the agent can inspect its own output.

The project targets people who use a coding agent and want consistent, presentable product-UI mockups rather than one-off prompts. The skills ship with the instructions the agent reads, so the quality of output depends on the agent following the bundled guidance.

Install and use

Prerequisites: Node.js with npx available, a coding agent that can load skills, and an existing event or project where the agent will run.

1. From your project directory, add the skill:

   npx skills add benjitaylor/skills --skill stageshow
   

2. Point your agent at the newly installed skill and ask it to build the UI you want. The skill's own instructions cover layout, motion, and the rendering step it uses to check its work.

3. Inspect the generated React output directly, or use the skill's filming tools to capture the result.

Results come from the agent's run; the repository does not include a standalone app that renders the demos for you. For the background and design rationale, see How it works; for browsable sample output, see the Stageshow preview.

Constraints and licensing

The skills are built for coding agents that support skill loading through the skills CLI; no other integration path is documented. Stageshow specifically assumes a React-based workflow and needs the agent to execute the bundled rendering and filming tools to verify output, so agents without shell or browser tooling may not complete that loop.

The repository does not declare a top-level license. Each skill includes its own licence and any third-party asset notices, so check the relevant skill before redistribution. The project does not describe platform support beyond the npx command.

HotRepo's take

The repository was new on 2026-11-01 and appeared in HotRepo's new-project list that day with 113 stars gained during a roughly six-hour window; no GitHub daily rank has been recorded for it. Its star count rose from 100 to 213 between 03:45 and 09:45 UTC that day, all within a single day of tracking, so the curve is too short to describe as a trend.

The project's appeal is narrow and concrete: it packages design and motion taste as agent-readable instructions rather than as a component library. That makes it easy to evaluate from the examples and skill file, but it also means output quality is tied to how well each agent follows instructions, which the repository's own materials do not measure across agents.