Agent Native

Collections for real jobs

Each collection is a short set of skills for one job, picked from the listed skills by a written rule. 10 collections so far.

Install two or three skills for the job at hand. Agents work better with a few focused skills than with many.

Collections with listed skills

  • Evals for agent apps

    Skills that help an agent write, run and read evaluations of LLM features and agents.

    24 listed skills

  • Payments with Stripe

    Stripe's own skills and others for checkout, subscriptions and billing work with Stripe.

    24 listed skills

  • Supabase, auth and row-level security

    Skills for building on Supabase: Postgres, auth, row-level security policies and edge functions.

    15 listed skills

  • Building and securing MCP servers

    Skills for writing, testing and locking down Model Context Protocol servers.

    24 listed skills

  • Browser automation

    Skills that let an agent drive a browser: Playwright, Puppeteer, browser agents and scraping.

    24 listed skills

  • Code review

    Skills that make an agent review diffs and pull requests the way a careful senior engineer would.

    13 listed skills

  • Testing and TDD

    Skills for test-driven development, unit and end-to-end tests, and fixing flaky suites.

    24 listed skills

  • Next.js and Vercel

    Vercel's own skills and others for Next.js apps, React Server Components and deploys on Vercel.

    24 listed skills

  • Observability and incidents

    Skills for logs, traces, error tracking and incident response with Sentry, Datadog, Grafana and others.

    24 listed skills

  • Security reviews

    Skills for security review, vulnerability scanning, secrets and threat modelling.

    24 listed skills

The skills listed here are third-party software, collected from public GitHub repositories as a catalog of recent skills and provided as is. Agent Native does not write or control them and is not responsible for any harm, loss or damage caused by installing or using them. Labels and checks describe what we saw at one commit and are not a guarantee. Read a skill's files before you run it. Terms of use