Why these skills

Instructions that survive contact with a real codebase

Each skill is written to be picked up at the right moment and followed without hand-holding.

  • They fire when they should

    A skill is only useful if the agent reaches for it unprompted. Descriptions are written and tested for exactly that.

  • Narrow on purpose

    One skill, one job. A skill that tries to cover everything gets loaded for everything and helps with nothing.

  • Open source, Apache-2.0

    Read them, fork them, adapt them to your own house style. Nothing here is a black box.

  • Two ways in

    Add the whole marketplace to Claude Code, or install a single skill from npm under the @kensio scope.

  • Versioned and changelogged

    Each plugin carries its own version, so an upgrade is something you choose rather than something that happens to you.

  • Written from real use

    Every skill here came out of work we actually do, not from guessing at what an agent might want.

Where they help

Built for the work that repeats

The tasks worth packaging are the ones you would otherwise explain again every week.

House style, enforced

Encode how your team writes tests, names things, and structures modules - once - and let the agent apply it.

Multi-step workflows

Release steps, migrations, review checklists: the procedures where the order matters and skipping one bites later.

Test data and fixtures

Factories and builders the agent can reach for instead of inventing a fresh shape every time.

Domain-specific tooling

Skills that carry the context an agent cannot infer from the code alone.

Add them to your setup

One command adds the marketplace. Pick what you want from there.