A grab bag of agent skills the Amika team has found useful. They aren't a coherent product, just tools we built for our own work and figured others might want too. Each skill lives in its own folder under skills/, and they all ship together as one Claude Code plugin.
Made with love by Amika. Join us on Discord to chat about software factories and agentic software.
Add this plugin to your Claude Code configuration:
{
"plugins": ["github:gofixpoint/agent-plugins"]
}Skills are invoked as /amika-scraper:<skill>, using the plugin name from .claude-plugin/plugin.json.
| Skill | What it does | Requires |
|---|---|---|
scrape |
Navigates a site with Playwright, records the interactions, then generates a standalone deterministic Playwright scraper in TypeScript. | Playwright MCP server |
video-gen |
Makes marketing graphics and short looping videos (PNG stills, MP4) for posts, launches, and decks. Today it has one style: isometric technical diagrams. | See the style's GUIDE.md |
/amika-scraper:scrape <url> <query>
For example:
/amika-scraper:scrape https://example.com/products Extract all product names and prices
The scraper works in two phases:
- AI-powered navigation. Claude drives the target site through the Playwright MCP browser to fulfill your query, recording a structured trace of every action (clicks, fills, scrolls, etc.) plus a narrative of the scraping strategy.
- Deterministic scraper generation. From the traces and narrative, Claude generates a standalone TypeScript Playwright script that reproduces the flow. The generated scraper has no AI/LLM dependencies, just Playwright.
- Create
skills/<name>/SKILL.mdwithnameanddescriptionfrontmatter, plus any supporting files the skill needs in the same folder. - Add a row to the Skills table above, and a short usage section if the skill takes arguments or has prerequisites.
MIT