:: CldKit AI Curator ::

ASK CLDKIT FIRST.
FIND SKILLS THAT MAKE YOUR AGENT BETTER.

Describe the job, stack, or error. CldKit starts with Skills, then adds MCP Servers and Recipes only when they help.

Ask CldKit

Works with mainstream agents

Reuse the same Skills across agents

Claude Code
Codex
Gemini CLI
Amp
Factory Droid
Grok Build
OpenCode
Pi

Evidence method

TRUST LABELS SHOULD BE AUDITABLE, NOT DECORATIVE.

CldKit separates review quality, runtime verification, and client compatibility so one signal never impersonates another.

01

VERSION-BOUND

Every review and runtime snapshot names the version it covers. A newer release creates visible drift.

02

RUNTIME-CHECKED

For supported MCP packages, we start the real process and record whether it initializes and which tools it exposes.

03

CLIENT-SPECIFIC

Claude Code, Codex, Cursor, and other targets are marked only after tested or caveated evidence exists. Untested stays unknown.

04

DRIFT STAYS VISIBLE

Expired checks, version changes, setup requirements, and failures remain visible instead of being flattened into a generic score.

CLI

INSTALL FROM TERMINAL

Search, verify, and install skills from your terminal

Every install is score-checked against our CldKit review database.

All commands

$ npx cldkit search testing

$ npx cldkit info playwright-testing

$ npx cldkit install playwright-testing

$ npx cldkit verify "npx -y @scope/pkg"

$ npx cldkit recommend

KNOW A GOOD COMPONENT?

Submit it for review. Help the community discover reliable agent tools.

SUBMIT COMPONENT