agent-toolkit AI Skill

What it is

agent-toolkit is a Skill collection from softaworks; the README says it's shared by leonardocouy to make daily work with Claude Code more efficient. It groups many Skills by category, plus some agents and slash commands, and each one installs as its own plugin. We sampled a few: agent-md-refactor splits a bloated AGENTS.md or CLAUDE.md into organized, linked files; codex has Claude Code call OpenAI's Codex CLI for code analysis, read-only sandbox by default; datadog-cli helps debug with Datadog logs and metrics. We didn't read the other items one by one. It follows the Agent Skills format, and the README says it also works in Codex, Cursor and similar tools.

Highlights:

Every Skill, agent and command installs separately
categories include documentation, design and frontend, planning, workplace communication, testing and Git
you can also paste a SKILL.md into claude.ai.

How to use:

First run npx skills add softaworks/agent-toolkit, or in Claude Code run /plugin marketplace add softaworks/agent-toolkit and then install only what you need, for example /plugin install humanizer@agent-toolkit.
Then ask for the task as each Skill describes.

Who it's for:

Developers who use Claude Code or a similar coding agent every day and want ready-made workflows to pick from.

Before you start:

You need Claude Code or another tool that supports Agent Skills.
Some items connect to outside services: codex needs the Codex CLI installed, and datadog-cli needs your own Datadog API key and then reads logs and metrics from your account.
Each service's costs are separate.
MIT license.

Based on the project's public documentation. Not tested by AI Inspiration World.