OpenClaw discovery layer

OpenClaw 火了之后，真正缺的是一个能帮你选 skill、看懂 skill、再安装的入口。

Install Agent Skills 不只是收集 GitHub 仓库。它应该帮你回答三件事：这个包到底是干什么的、适不适合你、值不值得现在装。

Start with trusted packages Browse all packages Read playbooks

Why people need this now

OpenClaw gave people a new install impulse.

People are not just asking “what is OpenClaw?” anymore. They are asking what to install first, what is actually useful, and how to avoid installing a random repo they do not understand.

This site should feel like a decision layer between hype and action: less vague catalog, more guided discovery.

Find packages that are actually install-shaped, not just vaguely AI-adjacent.

See what a package contains before copying the install command.

Compare trust, risk, and skill coverage without leaving the site.

What is different here

The site should optimize for install decisions, not just GitHub discovery. That means package intent, starter skill guidance, and visible risk context.

Who this is for

People arriving from the OpenClaw wave, AI agent tinkerers, and teams trying to standardize which packages are safe enough to try first.

What to do first

Start from the trusted shortlist if you want a cleaner path. Use the broader packages page when you already know the kind of workflow you need.

A better browse flow

Use the site in three steps

The goal is not to make you read every page. The goal is to move you from “OpenClaw is interesting” to “I know which package to try first and why”.

Install workflow

Copy the command only after you understand the package

The site should not rush people from hype to install. It should help them understand the package first, then make the install command obvious when they are ready.

Default command

One command, but not blind install

npx skills add agentworks/secure-skills

1. Shortlist a package

2. Review fit, starter skill, and risk

3. Then copy the install command

Browse packages•Trusted shortlist

Install on your agent

Docs

Claude Code skills

Create skills directory
Install a package (recommended)

mkdir -p ~/.claude/skills

Or install via: npx skills add owner/repo

Open guide

Codex skills

Install new skills (CLI or manual)
Restart Codex to pick up new skills

mkdir -p ~/.codex/skills

Then restart Codex.

Open guide

Use Agent Skills in VS Code

Store skills inside your repo
Share with your team

.github/skills/my-skill/SKILL.md

Open guide

Cursor agent skills

Add a remote rule from GitHub
Apply skills to the project

npx skills add owner/repo

Open guide

FAQ

What should I do before installing a package?

Read the package detail page first: check what the package is for, which skill to start with, and whether the risk notes match your tolerance.

Why is this site useful for the OpenClaw wave?

Because people need more than a GitHub link. They need help deciding which packages are installable, understandable, and worth trying first.

What is the default install command?

Run npx skills add owner/repo to install a package in seconds. Use the package detail page to decide whether you should run it.

Step 01

Start from trusted packages

If you are new to the ecosystem, do not start from the noisiest or newest repo. Start from the shortlist that looks installable and understandable.

Open trusted shortlist

Step 02

Read the package like a product

A useful detail page should tell you what the package is for, when to skip it, which skill to read first, and how much real implementation it bundles.

Browse package details

Step 03

Install only after you understand it

The install command should be the last step in the decision flow, not the first thing you see before you know what the package actually contains.

See install flow

Recommended first click

Trusted packages to start with

Best for security-sensitive workflows

agentworks/secure-skills

低风险

Curated, low-risk skills for code review, research, and triage.

可信度 86/1003 个技能Marketplacegood first installsecurity-sensitive

Signals

Stars 1,280 · 更新于 2026年1月10日

Topics

code-review · security · agents

查看

Best for operations and runbooks

studio-nova/ops-skillbox

中风险

Operations automation skills with clear guardrails and install docs.

可信度 63/1002 个技能ops workflowsecurity-sensitive

Signals

Stars 540 · 更新于 2025年12月22日

Topics

devops · automation

查看

Best for reusable agent workflows

openai/openai-agents-python

中风险

A lightweight, powerful framework for multi-agent workflows

可信度 46/1001 个技能

Signals

Stars 18,567 · 更新于 2026年1月23日

Topics

agents · ai · framework

查看

Safer entry points

Low-risk packages for first-time installs

Best for security-sensitive workflows

agentworks/secure-skills

低风险

Curated, low-risk skills for code review, research, and triage.

可信度 86/1003 个技能Marketplacegood first installsecurity-sensitive

Signals

Stars 1,280 · 更新于 2026年1月10日

Topics

code-review · security · agents

查看

Fresh packages worth evaluating

Best for MCP and agent tooling

lobehub/lobehub

中风险

The ultimate space for work and life — to find, build, and collaborate with agent teammates that grow with you. We are taking agent harness to the next level — enabling multi-agen...

可信度 46/1001 个技能coding workflow

Signals

Stars 73,543 · 更新于 2026年3月13日

Topics

agent · agent-collaboration · agent-harness

查看

Best for reusable agent workflows

VoltAgent/awesome-openclaw-skills

中风险

The awesome collection of OpenClaw skills. 5,400+ skills filtered and categorized from the official OpenClaw Skills Registry.🦞

可信度 46/1001 个技能

Signals