Awesome Claude Skills (Composio)
ComposioHQ/awesome-claude-skillsA curated list of Claude skills plus hundreds of templated app-automation skills that run through Composio's hosted Rube MCP.
With caveats
Broad access Depth 1/5
anbeime Skill Store Claude skills
A Chinese-language skill store: 63 local skills (content, video, e-commerce, PPT) plus daily-synced mirrors of other authors' skills.
Our verdict Skip it
A mirror-plus-grab-bag with an unclear license, a chat-persona root skill and bundled copies of other authors' work; install the originals or the specific local skill you want instead.
Depth expertise shipped Security Broad access Maintenance Updated today Adoption stars License no clear license
The author changed 1 of 93 skill folders after our review (noticed Oct 29, 2026). The review describes the earlier version; we recheck changed skills weekly, the most significant changes first.
What anbeime Skill Store does
This repository is a collection rather than a single skill. The README (mostly Chinese) says it holds 182 "official" skills crawled daily from six upstream repositories (including OpenAI skills, VoltAgent lists and Matt Pocock's skills) and 63 local skills grouped by theme: content creation and publishing, video, e-commerce and marketing, PPT, voice and audio, digital avatars, and documents and analysis. The listed local skills include WeChat and X posting helpers, a Xiaohongshu image generator, a 1688 cross-border e-commerce pipeline, an SRT-to-whiteboard-animation video skill, TTS and ASR skills, and a set of "antinet" document skills that parse documents and produce four-color fact/explanation/risk/action cards.
The root SKILL.md is not a catalog guide but a virtual companion persona (xiaoyue-companion) that calls Zhipu's GLM model and sends messages through OpenClaw. The tree has roughly 2,300 files, with Python and Node scripts, Windows scheduler setup and a Vercel config alongside the skills. Quality varies by skill, and some copies of third-party skills are bundled. The README also promotes a separate paid knowledge-assistant product with an email sign-up. The license field is empty; the README mentions both CC BY 4.0 and MIT.
Good for
- Chinese-language content, video and e-commerce workflows
- Browsing a large set of skills to find ideas to adapt
- OpenClaw or Hermes users who want local Chinese skills
Not for
- Anyone who needs clear licensing for commercial reuse
- Users who want a single focused, maintained skill
What's inside depth 2/5
93 skill folders averaging about 9.6 KB with many scripts overall, but uneven quality, little structure shared between them and no evals.
- Skills
- 93
- Reference docs
- 384
- Scripts
- 591
- Tests / evals
- 103
- Templates
- 285
- Avg SKILL.md
- 9.4 KB
Counted from the repository file tree. How we score depth.
Dependencies and costs
- Zhipu AI API key for the companion skill
- Various third-party APIs used by individual skills (TTS, image, video, 1688)
- Node.js and Python
- Chrome automation skill clones and builds vercel-labs/agent-browser
Security review
Broad accessThis skill asks for more access than most, or sends data to another service. That is usually a design choice rather than bad intent, but you should know before you install it.
- Pre-approves broad shell commandsFrontmatter
allowed-tools: Bash(*)pre-approves unrestricted shell for the chrome-automation skill.skills/chrome-automation/chrome-automation/SKILL.md:4 - Pre-approves broad shell commandsThe xiaoyue-companion skill (and projects/companion-simple/SKILL.md) pre-approves
Bash(node:*) Bash(npm:*) Bash(openclaw:*) Bash(curl:*), wildcards over interpreters and curl.SKILL.md:4 - Uses your browser cookies or logged-in profileOn first run rsyncs your real Chrome profile (login states, cookies, settings) into a Chrome-Automation profile, launches Chrome with --remote-debugging-port=9222 and drives it over CDP; the Windows script does the same.skills/chrome-automation/chrome-automation/scripts/start-chrome-mac.sh:24
- Turns off permission prompts or safety checksTells the agent to run the installer check before use and says it installs dependencies automatically 'without user confirmation'.skills/chrome-automation/chrome-automation/SKILL.md:11
- Loads instructions from a server at run timeTells the agent to add a periodic step that fetches https://www.moltbook.com/heartbeat.md and follow it every 4+ hours, so behaviour is controlled by the remote server.skills/moltbook/SKILL.md:170
A large mixed collection: a companion chat skill, Chinese content and e-commerce skills, a local email assistant, social-publishing scripts and a Chrome automation skill. chrome-automation pre-approves any shell command, installs a toolchain without asking and copies your real Chrome profile (including cookies and logins) into a debugging-enabled automation profile. The moltbook skill instructs the agent to fetch and follow a remote heartbeat file every few hours, and the email assistant keeps your IMAP/SMTP app password in a local config file.
Worth knowing
- Installs with curl | shInstalls Rust with
curl ... https://sh.rustup.rs | sh -s -- -ywhen cargo is missing; scripts/install-scripts.md line 264 also showscurl ... vercel-labs/agent-browser/main/install.sh | bash.skills/chrome-automation/chrome-automation/scripts/auto-install-mac.sh:68 - Runs remote code at an unpinned versionClones the default branch of github.com/vercel-labs/agent-browser, runs pnpm install,
npx playwright install chromiumand builds the native binary, all unpinned.skills/chrome-automation/chrome-automation/scripts/auto-install-mac.sh:20 - Installs packages or toolsemail-concierge sets up a virtualenv with pip packages (openvino-genai, optimum, sounddevice and others) and downloads an ~2 GB local model from ModelScope.skills/email-concierge/setup_auto.py:1
- Makes network requestsSkills call third-party APIs: Zhipu (ZHIPU_API_KEY) for chat and images, WeChat and X publishing through scripted Chrome, 1688/TTS/video generation services, Moltbook, and the user's IMAP/SMTP mail server.projects/companion-skill/src/image-generator.ts:73
- Needs an API key for its own serviceReads ZHIPU_API_KEY (and WECHAT_APP_ID/WECHAT_APP_SECRET from environment or ~/.baoyu-skills/.env) and uses them only against their own service; email-concierge stores IMAP/SMTP app passwords in a local JSON config.scripts/xiaoyue-chat.js:7
- Contains compiled binariesThe repo contains 11 zip/binary files (opspilot demo, worker_packages/*.zip) that we cannot read.antinet-agentteams/opspilot-zero-demo.zip:1
What we checked
The repository (2299 files) on Oct 29, 2026. Not a full review: 1506 of 2299 files were not downloaded (large repository; we reviewed the skill folders); it contains compiled binaries or archives we cannot read; we read 3 of 128 code files; it relies on code outside the repository: 1,506 of 2,299 files were not downloaded, including antinet-agentteams/core/*.py and most worker packages; 11 binary/zip files (opspilot-zero-demo.zip, worker_packages/*.zip and others); vercel-labs/agent-browser repository cloned and built by chrome-automation, plus Rust toolchain, pnpm and Playwright Chromium.
NVIDIA SkillSpector static scan: 440 pattern hits (MCP Rug Pull 125, Privilege Escalation 93, Agent Snooping 64, Supply Chain 57, Data Exfiltration 29, Tool Misuse 24). Most of the 440 hits come from the repo being a scraped catalog of other skills (a Burp Suite entry in data/skills.json, README link lists, 'dsh-*' plugin names) and from documentation (DELETE routes, .env and ssh key mentions). The hits that matched real behaviour are recorded as flags: Bash(*) and Bash(curl:*) allowed-tools, the Chrome profile import, rustup and agent-browser installs, and the moltbook heartbeat fetch.
A review, not a certification. Skills change after we check them — read the files before installing.
License
No license file (GitHub reports none) — the README badge says CC BY 4.0 and another line says MIT; bundled third-party skills carry their own licenses, so treat reuse as unclear.
Can it run unattended?
Not really. The repo has a 24-hour sync crawler, but its skills are mostly interactive content tools without stop conditions or verification.
Alternatives to anbeime Skill Store
Agentic Awesome Skills (AAS)
sickn33/agentic-awesome-skillsA catalog of roughly 2,900 community SKILL.md playbooks, with a local CLI, MCP server and plugin bundles for searching and installing them.
With caveats
Check first Depth 3/5
daymade Claude Code Skills
daymade/claude-code-skillsA marketplace of about 120 skills from one author: skill-creator fork, audio and transcription tools, Claude Code and Codex ops, research, QA and more.
With caveats
Broad access Depth 4/5
md2wechat
geekjourneyx/md2wechat-skillA CLI plus agent skill that converts Markdown into WeChat Official Account HTML, makes covers and images, and pushes drafts to the WeChat draft box.
With caveats
Check first Depth 3/5
FAQ
What is anbeime/skill?
It is a Chinese-language skill store that combines locally written skills for content, video and e-commerce with daily-synced mirrors of skills from other repositories. It is a collection, not a single skill.
How do I install anbeime/skill?
The repo has a plugin manifest, but the better route is to copy only the individual skill folders you need into your agent's skills directory. Installing everything adds 93 skill folders and a root persona skill that you probably do not want.
Is anbeime/skill safe?
We saw no hidden payloads in the files we read, but it is a large mix of scripts from several authors, including browser automation and publishing helpers. Review each skill before enabling it.
Can I use these skills commercially?
The license is unclear: GitHub detects none, and the README mentions both CC BY 4.0 and MIT. Bundled skills from other authors keep their own licenses, so check the upstream repos.
Sources: GitHub repository (stars, license, file tree, last push Oct 29, 2026). Review based on the README and SKILL.md files as published.