# AgentSkillsHub CLI
**[🌐 Website](https://agentskillshub.top)** · **[🔎 Browse Skills](https://agentskillshub.top)** · **[🛡️ Security Report](https://agentskillshub.top/blog/securing-117k-ai-skills/)** · **[🏢 Enterprise](https://agentskillshub.top/enterprise/)**
Search, audit, and install open-source AI agent skills & MCP servers from your terminal. Every result is **security-graded** and **quality-scored** by [AgentSkillsHub](https://agentskillshub.top) — a directory of 117K+ skills with a trust signal before you install.
```bash
npx @agentskillshub/cli search "scrape a website" --safe
```
## Why
Discovering a skill is easy. Knowing whether it's safe to run against your credentials is not. It puts the trust signal *before* the install:
```
$ npx @agentskillshub/cli search postgres --category mcp-server --limit 2
call518/MCP-PostgreSQL-Ops 150★ 🟢 SAFE
Give AI assistants full PostgreSQL DBA superpowers — 30+ tools…
mcp-server · claude-code · mcp · python https://agentskillshub.top/skill/call518/MCP-PostgreSQL-Ops/
sgaunet/postgresql-mcp 5★ ⚪ UNAUDITED ~23.0k tok
A Model Context Protocol (MCP) server… read-only query execution…
mcp-server · claude-code · go · mcp https://agentskillshub.top/skill/sgaunet/postgresql-mcp/
```
## Commands
Run via `npx @agentskillshub/cli <command>`:
| | |
|---|---|
| `search <query> [filters]` | Find skills. Filters: `--category` `--platform` `--min-stars` `--safe` `--limit` |
| `audit <owner/repo>` | Free basic trust check: security grade + plain-English verdict |
| `install <owner/repo>` | Install commands + "check before you install" safety line |
| `update` | Force-refresh the cached index |
Add `--json` to any of `search` / `audit` / `install` for machine-readable output.
## How it works
The catalog (~20K quality skills, stars ≥ 5) is a single static file (~1.7MB gzipped) served from the CDN. The CLI downloads it once, caches it at `~/.cache/agentskillshub/`, and checks for a newer index with a cheap 77B probe (re-downloading the full file only when it actually changed, so a freshly-deployed index reaches you within minutes). **All searching is local** — fast, works offline, and puts zero load on the backend.
## Security grades
🟢 SAFE · 🟡 CAUTION · 🔴 UNSAFE · ⛔ REJECT · ⚪ UNAUDITED
⚪ **UNAUDITED** is not "probably fine" — it means *no one has audited it*. Check the code, the credentials it asks for, and who maintains it before you trust it.
We scanned the whole catalog and wrote up what we found — only 17.7% of skills are popular enough to be graded, 1 in 32 graded ones is unsafe, and the risk lives in the long tail: **[We security-graded 117,854 AI agent skills](https://agentskillshub.top/blog/securing-117k-ai-skills/)**.
## Free vs. Pro
- **Free**: search · basic audit · install commands, for any catalogued skill.
- **Pro members**: set `ASH_PRO_KEY` and `search` matches the full README text (up to 200 results). Each search has a time limit; when it stops there, the CLI says the list may be incomplete and how to narrow it, and `--json` rows carry `incomplete_results: true`.
- **Pro / Enterprise**: 5-dimension deep audit, any GitHub URL (incl. <5★ / private), CI/batch auditing, compliance evidence → <https://agentskillshub.top/enterprise/>
## Env
| Var | Default |
|---|---|
| `AGENTSKILLSHUB_BASE` | `https://agentskillshub.top` |
| `AGENTSKILLSHUB_CACHE` | `~/.cache/agentskillshub` |
| `ASH_PRO_KEY` | unset — Pro member key for README-depth search (get one at <https://agentskillshub.top/pro/>) |
| `NO_COLOR` | unset (set to disable color) |
MIT © AgentSkillsHub