5 Must‑Try AI Skills: Find One That Fits or Build Your Own

Many readers ask the author for a list of "must‑install" AI Skills. The author argues that a Skill’s value depends on how well it matches a specific workflow, and proposes a simple approach: first search for a suitable Skill, and if none exists, create one yourself, gradually building a personal Skills repository.

1. find-skills – Learn to "find a Skill"

This entry‑point Skill helps you locate the right Skill for a given need. By describing the desired functionality to an AI, it retrieves matching Skills and can install them automatically.

I want a Skill that can organize meeting minutes.

Install with:

npx skills add https://github.com/vercel-labs/skills --skill find-skills

URL: https://skills.sh/vercel-labs/skills/find-skills

2. skill-creator – When you can’t find one, build it

skill-creator assists you in turning personal ideas, data, and workflows into a reusable Skill, even if you are not familiar with the full specification format. It is especially useful for private‑assistant tasks such as article summarization, draft checking, or meeting‑recording processing.

Install with:

npx skills add https://github.com/anthropics/skills --skill skill-creator

URL: https://skills.sh/anthropics/skills/skill-creator

Suggested starter scenarios:

Automatically format reading notes after each article.

Check titles, introductions, AI‑tone, and factual‑check points in each draft.

Generate meeting minutes, to‑do items, and shareable summaries from recordings.

3. skill-vetter – Security review before installation

skill-vetter performs a pre‑install audit of Skills sourced from OpenClaw/ClawHub/GitHub. It checks for suspicious metadata, excessive permissions, network + shell requirements, references to secrets, risky commands (curl, wget, nc), Base64 obfuscation, prompt injection, typosquatting, etc., and outputs a safety rating (SAFE / WARNING / DANGER / BLOCK).

Install with:

npx skills add https://github.com/useai-pro/openclaw-skills-security --skill skill-vetter

URL: https://skills.sh/useai-pro/openclaw-skills-security/skill-vetter

4. dokobot – A helper for reading web content

dokobot solves the common bottleneck of feeding web material into AI workflows. It extracts page content (including behind‑login pages) and passes it to AI for summarization, translation, rewriting, or organization. The browser plugin can reuse login sessions for sites that block simple crawlers.

Key use cases:

Summarize and rewrite web articles.

Collect documents from web pages for later processing.

Fetch content from platforms such as Twitter, YouTube, WeChat, Weibo, Zhihu, Xiaohongshu, Bilibili.

Search engines like Google, Bing, DuckDuckGo, X, Baidu, Sogou.

URL: https://dokobot.ai/skill

5. superpowers – A coding‑focused Skills suite

Designed for developers who use AI coding assistants (Claude Code, Codex, Cursor, Qoder), superpowers bundles multiple Skills that cover brainstorming, planning, execution, code review, debugging, and parallel development.

GitHub: https://github.com/obra/superpowers

Target audience:

Programmers who frequently use AI coding tools.

People who want AI to break down tasks before execution.

Those who want AI to participate in the full development lifecycle, not just autocomplete.

Teams experimenting with multi‑agent parallel work but lacking methodology.

The author cautions that adopting such Coding Skills changes the collaboration rhythm; initial speed gains may be modest as users adapt to a "plan‑then‑execute‑then‑review" workflow, but the long‑term benefit is a more stable AI‑augmented development process.

Further resources for discovering Skills

skills.sh – a beginner‑friendly portal with many installable Skills (https://skills.sh/).

skillsmp – useful for searching, categorizing, and observing Skill trends (https://skillsmp.com/).

The author’s public Skills repository on GitHub (https://github.com/chujianyun/skills), which includes Skills for optimization, Agent design, Claude.md, source‑code analysis, paper reading, OpenClaw, Hermes operations, etc.

The repository is also indexed by the Meituan‑hosted site xia345.com.

Conclusion

The most useful Skills are those that fit your actual scenario. AI tools and models evolve, but a well‑documented workflow, experience, and data that you continuously refine become lasting productivity assets. Start with a concrete small task, install a Skill that solves it, and if none exists, build one yourself, gradually expanding your personal Skills toolbox.