Install
$ agentstack add mcp-xm1k3-ai-community-skills ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
ai-community-skills
All the community Agent Skills scattered across GitHub, in one local catalog. Search them, browse them, and install them into Claude Code, Codex, or Grok, with a risk check on every skill as a bonus.
- [Sources](#sources)
- Getting started
- [Installation](#installation)
- [Commands](#commands)
- [Usage](#usage)
- [Web dashboard](#web-dashboard)
- [Risk analysis](#risk-analysis)
- Reference
- [Config](#config)
- [Updates](#updates)
- [Disclaimer](#disclaimer)
- [License](#license)
Sources
The curated list of community repositories lives in [config.json](config.json) at the root of this repository. acs init downloads the latest version of that file and writes it to ~/.acs/config.json, so a fresh install always starts from the current list (pass --offline to use the copy bundled with the package instead).
| Source | Repository | Enabled | | --- | --- | --- | | anthropic-skills | anthropics/skills | yes | | superpowers | obra/superpowers | yes | | composio-awesome-claude-skills | ComposioHQ/awesome-claude-skills | yes | | sickn33-agentic-awesome-skills | sickn33/agentic-awesome-skills | yes | | alirezarezvani-claude-skills | alirezarezvani/claude-skills | yes | | behisecc-awesome-claude-skills | BehiSecc/awesome-claude-skills | yes | | travisvn-awesome-claude-skills | travisvn/awesome-claude-skills | yes | | humanlayer-skills | humanlayer/skills | yes | | mattpocock-skills | mattpocock/skills | yes | | k-dense-scientific-agent-skills | K-Dense-AI/scientific-agent-skills | yes |
Know a good repository of skills that is missing? Open a pull request that adds it to config.json:
{ "name": "owner-repo", "repo": "https://github.com/owner/repo", "enabled": true }
name is the folder the repository is cloned into and the label used everywhere in the catalog, repo accepts https URLs, owner/name GitHub shorthand, ssh URLs, or local file:// URLs. trust (0-100, default 50) says how much the copy from that repository should win when the same skill exists in more than one source.
Installation
npm install -g ai-community-skills
Or run it without installing anything:
npx ai-community-skills init
The binary is available as ai-community-skills and as the short alias acs. Node.js 20 or newer and git on PATH are required. After installing, run acs init to write the default config to ~/.acs/config.json, then acs sync to build the catalog.
Commands
| Command | Description | | --- | --- | | acs init | Download the curated source list and write ~/.acs/config.json. --offline uses the bundled copy. | | acs sync | Clone or update every enabled source and rebuild the index. | | acs list | List indexed skills with filters for category, risk, source, and findings. | | acs search | Search skills by name, description, category, and tags. | | acs info | Full metadata, risk flags, findings, source reputation, and install state. | | acs validate [skill] | Check frontmatter, referenced paths, and description quality. | | acs install | Install a skill after a risk summary and a confirmation. | | acs uninstall | Remove an installed skill. | | acs dedupe | Remove duplicate copies of the same skill across sources. | | acs stats | Counts by category, source, risk level, and finding. | | acs export-awesome-list | Generate an awesome-list style markdown file from the catalog. | | acs ui | Start the local web dashboard on port 8080. |
Every command that writes to disk has a --dry-run flag or asks for confirmation unless --yes is passed. Run acs --help for the full list of flags.
Usage
Build the catalog:
$ acs sync
Syncing anthropic-skills (https://github.com/anthropics/skills) ... updated, 19 skills indexed, 1 skipped
Syncing superpowers (https://github.com/obra/superpowers) ... unchanged, 14 skills kept
...
Index rebuilt with 3315 skills from 5 source(s).
A source whose repository did not change since the last sync keeps its index entries instead of being rescanned, so a routine sync takes a few seconds. The index is rebuilt anyway after upgrading acs, or on demand with acs sync --force.
Search and inspect:
$ acs search "frontend design"
$ acs info frontend-design
Install into Claude Code (personal scope, ~/.claude/skills/), into the current project, into Codex, or into Grok Build:
$ acs install frontend-design
$ acs install frontend-design --project
$ acs install frontend-design --target codex
$ acs install frontend-design --target grok
$ acs install frontend-design --target web
--target web produces a zip in ./acs-exports/ for manual upload to claude.ai. When a skill name exists in more than one source, add --source . Installed copies are never updated automatically: acs sync warns when the upstream copy changed, and you decide whether to reinstall.
Web dashboard
acs ui
Opens a local dashboard on http://127.0.0.1:8080 with the same catalog: search and filters, skill pages with rendered SKILL.md and a file browser, every risk finding with a link to the exact line, favorites and groups that can be installed in one go and shared as JSON, source management with a trust level per repository, and duplicate cleanup. Every view has a shareable URL.
Risk analysis
Every skill is analyzed statically, nothing found in a skill is ever executed. Each entry records whether it ships scripts, references network calls, contains destructive operations and whether they are paired with a confirmation, uses Claude Code specific frontmatter, contains prompt injection patterns, or references secrets and credential files. The result is a low, medium, or high risk level plus the list of findings, each with file, line, and excerpt, so you can judge false positives yourself before installing.
Duplicates across sources are detected with a normalized content hash that ignores catalog metadata and whitespace, and the copy from the most trusted source is kept.
Config
Your local ~/.acs/config.json is yours to edit, from the file or from the Sources page of the dashboard:
{
"sources": [
{ "name": "anthropic-skills", "repo": "https://github.com/anthropics/skills", "enabled": true, "trust": 100 },
...
],
"dedupeAfterSync": true
}
All data lives under ~/.acs/ (config, index, install records, favorites and groups, cloned sources). Set ACS_HOME to use a different directory.
Updates
Once a day (once an hour while acs ui is running) acs asks the npm registry for the latest published version. When a newer one exists, every command prints a short notice on stderr and the dashboard footer shows an upgrade button with the command to run:
npm install -g ai-community-skills@latest
The check never blocks a command for more than a moment, the result is cached in ~/.acs/update-check.json, and nothing else is sent. Set ACS_NO_UPDATE_CHECK=1 to turn it off.
Disclaimer
Skills are community content pulled from third-party repositories. The risk analysis is heuristic and static: it will produce false positives and can miss real problems. Read a skill before installing it, especially anything rated high. The author is not responsible for what an installed skill does in your environment.
License
ai-community-skills is distributed under the MIT License.
Source & license
This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: xm1k3
- Source: xm1k3/ai-community-skills
- License: MIT
- Homepage: https://xm1k3.github.io/ai-community-skills/
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.