English · 中文

skillzoo

Drop-in skills & slash-commands for AI coding agents (Claude Code, Codex, Kimi, GLM, …). You don't install them by hand — you hand one prompt to your agent and it sets everything up.

A small, growing collection of agent capabilities I find genuinely useful, packaged so anyone can add them in ~30 seconds.

---

⚡ Install in ~30 seconds (your agent does the work)

You do not need to know git. You need two things: (1) be signed in to GitHub, (2) paste one prompt to your agent.

1. Copy this repo's URL — the green <> Code button at the top of this page → HTTPS → copy the git clone https://github.com/zhenzoo/skillzoo.git line. (Or just copy the URL from your browser's address bar.)
2. Paste this whole prompt to your agent (Claude Code / Codex / Kimi / GLM / Claude / …):

Clone https://github.com/zhenzoo/skillzoo.git
(I'm signed in to GitHub via the `gh` CLI — if I'm not, walk me through `gh auth login` first.)

From that repo, install the `intuitive-explain` command at the USER level of my agent — this is
general infrastructure I want everywhere, NOT project-specific:
- Claude Code: copy commands/intuitive-explain.md into ~/.claude/commands/  (or my CLAUDE_CONFIG_DIR/commands/).
- Other agents (Codex / Kimi / GLM / …): put it in the equivalent USER-level commands/prompts
  folder; adapt the file's format if that agent needs it, but keep the instructions intact.

Do NOT ask me where to put it — user level is correct, just do it.

When you're done, report back so I feel it immediately:
(1) what /intuitive-explain does, (2) how I trigger it (I type `/intuitive-explain <thing>`),
(3) when it's worth using, and (4) ONE concrete example of its output.

3. That's it. If you're signed in to GitHub, your agent clones the repo, drops the file at user level, and reports back with a live example. No manual file juggling.

---

🧩 What's inside

Canonical list = skills.yaml (single source of truth — the table below just mirrors it).

id	type	what it does
intuitive-explain	command (slash)	A super-boost for your AI agent's explanations. Instead of dumping a wall of text, it lays out the real thing across many dimensions — a plain-language analogy + real files/objects it opens on your machine + visualizations + a draggable interactive page + side-by-side comparisons — so you build intuition and actually get it, not just skim text. Other AIs can send a table or a card, but too shallow (like an encyclopedia); this fills the "still don't get it + not multi-faceted" gap.

---

🎬 Examples

Real ways I use it — say "I don't get this, explain it intuitively" and it goes multi-dimensional instead of dumping text:

• Scrolled past a video with a concept you can't follow? Paste the link: /intuitive-explain what's this video actually about — don't just give me text → it downloads + transcribes + researches, then shows you the real thing. (It decoded an "IT ops vs ITSM" video for me: what a real ticketing UI looks like + GLPI, a free ITSM system you can install and poke at yourself.)

  

• Drowning in "embedding / vector / reranker"? /intuitive-explain embedding → an embedding is just a coordinate on a "semantic map"; it hands you Google's embedding projector so you drag a 10,000-word cloud and see similar words clump together — that's when it clicks.

  

• Not sure which tool to reach for? /intuitive-explain when do I use AnySearch vs Jina vs a real-browser agent → two clean groups, side by side.

  

• You're on your phone, away from the machine? It builds a draggable interactive page (an Artifact) and hands you a link — drag the slider and watch "estimate vs real" split apart. Local windows you can't see? Skipped — it meets you where you are.

  

Full walkthrough (in Chinese, with all the real screenshots): my Xiaohongshu post on this skill.

---

🔧 command vs skill — and where it goes

• command = one markdown file (e.g. commands/intuitive-explain.md). You trigger it manually by typing /intuitive-explain <thing>. intuitive-explain is a command.
• skill = a folder (skills/<name>/SKILL.md + optional scripts). The agent auto-invokes it when it detects the matching situation — you don't type anything.
• Install at USER level, not project level — these are general infrastructure you want in every project. (Claude Code: ~/.claude/commands/ and ~/.claude/skills/.)
• Some capabilities can ship as both (a command for manual trigger and a skill for auto-trigger). Each entry in skills.yaml says which it is and how it triggers.

---

➕ Adding a skill (maintainers)

1. Drop the file(s): commands/<name>.md (command) or skills/<name>/SKILL.md (skill).
2. Register it in skills.yaml — the single source of truth (id · type · path · trigger · summary).
3. No build step. No hardcoded machine paths. Keep each skill self-contained.

---

License

MIT — free to use, copy, adapt. Attribution welcome, not required.