Borrow skills. Stay light.

Borrow the right agent skill, then let it go.

Skillcat is a lightweight loader for Claude, Codex, and AI coding agents. Ask for what you need, let Skillcat fetch and adapt the right workflow, then keep your context clean.

“Find me a good Stripe webhook skill for this repo.”Skillcat pads in, grabs the skill, pads out

Find → adapt → use → keep or toss. No skill-hoarding required.

1tiny loader to install
on demandskills when the task needs them
lesscontext clutter

Different by design

Directories make you collect. Skillcat helps you actually use.

Most registries stop at discovery. Skillcat sits in the workflow and brings in specialized instructions only when they are actually useful.

01

Install Skillcat once

Your agent gets a small, polite loader — not a giant backpack of instructions.

02

Ask for the capability you need

Say what you want to do. Skillcat finds, fetches, and loads the right skill only for that moment.

03

Keep the good stuff

Save what works, toss what does not, and avoid rewriting the same workflow by hand.

Not another directory

Directories help you browse. Skillcat helps your agent use the right skill at runtime.

Anti-hoarding mode

Try an interesting skill without making it part of your permanent setup. Keep only the ones that earn their spot.

Centralized management

Organizations get a governed registry with versions, approvals, repo-specific adaptations, and auditability.

Install once

One tiny loader for Claude or Codex.

Instead of manually installing a pile of skills, install the loader. Describe the task, and Skillcat brings back the useful instructions while leaving the rest outside your context window.

Claude Codeclaude skill install skillcat/loader
OpenAI Codexcodex skill add skillcat/loader

For organizations

Write a skill once. Adapt it everywhere.

One skill often needs different behavior in different repos, teams, and environments. Skillcat provides a controlled registry layer: version skills, adapt them per repo, approve changes, and let every team download the right variant automatically.

Central registryPrivate skills, public mirrors, ownership, and access control.
Version controlPin stable versions, test upgrades, roll back when a workflow regresses.
Automatic adaptationRepo conventions, team rules, tools, and compliance notes applied for you.

Private beta

Cleaner agents, safer trials, fewer mystery folders.

We stopped stuffing every promising skill into Claude. Now our agents fetch the right workflow only when needed.
Maya ChenFounder, Buildflow
This is the missing layer between public skill chaos and what a real team can actually adopt.
Jon BellEngineering Lead
Write once, adapt per repo, keep versions controlled. That is exactly how agent skills should work in companies.
Priya RamanAI Ops Consultant