The Skills settings manage the reusable skills available to this project’s agents. Skills are reusable instruction bundles the agent can invoke on demand — a way to give agents repeatable, well-defined capabilities without bloating every prompt.
On this page
How skills work
Active skills are advertised in the agent’s system prompt so it knows they’re available, but their full contents load only when the agent actually calls invoke_skill. That keeps the agent’s context lean — it’s aware of what it can do without carrying every skill’s instructions until the moment one is needed.
Curated skills
The Curated section lists ready-to-use skills you can add with one click. Each shows a name, a short description, and an Install button; recommended ones are flagged. Examples include:
- Playwright CLI — browser automation and E2E testing via the Playwright CLI.
- Frontend Design — create distinctive, production-grade frontend interfaces with strong design fundamentals.
- Systematic Debugging — a disciplined debugging workflow: reproduce, isolate, hypothesize, verify.
- Test-Driven Development — a red-green-refactor workflow with failing tests written before implementation.
Installed skills
The Installed section shows the skills currently available to this project’s agents. Until you add one, it shows that no skills are installed for the project yet.
Installing from a registry
Beyond the curated list, you can install skills from a registry. Install from registry shells out to npx skills add to fetch and add a skill.
Installing from a registry installs third-party code. Only install skills from sources you trust.
Start with the Curated skills — they’re vetted and one click to add. Reach for Install from registry only when you need something specific that isn’t in the curated list, and confirm the source is trustworthy first.
