Contents

OpenClaw Must-Have Skills — Highly Recommended

The OpenClaw Skills ecosystem (ClawHub) has surpassed 13,000 skill packages, but quality varies wildly. A February 2026 security audit flagged ~13.4% for critical issues. This post recommends genuinely useful Skills by use case, helping you avoid the minefield.

Installation

All Skills are installed via the clawhub command:

# Install an official Skill
clawhub install openclaw/github

# Install a community Skill
clawhub install community/obsidian-claw

# View installed Skills
openclaw skills installed

# Search
openclaw skills search "keyword"

⚠️ Always check author reputation, star count, and source code before installing community Skills. 341 malicious Skills were found stealing user data on ClawHub.


🏆 Top 10 Must-Install Skills

Rank Skill Category Install Status
1 GitHub Development clawhub install openclaw/github 🟢 Stable
2 Web Browsing Research Built-in 🟢 Stable
3 GOG Productivity clawhub install openclaw/gog 🟢 Stable
4 Tavily Research clawhub install framix-team/openclaw-tavily 🟢 Stable
5 Gmail Productivity Built-in 🟢 Stable
6 Calendar Productivity Built-in 🟢 Stable
7 Slack Communication clawhub install steipete/slack 🟡 Beta
8 n8n Automation clawhub install community/n8n-openclaw 🟡 Beta
9 Obsidian Productivity clawhub install community/obsidian-claw 🟡 Beta
10 Home Assistant Smart Home clawhub install openclaw/homeassistant 🟡 Beta

🔧 Developer Essentials

GitHub (#1, Score 72/80)

The highest-rated Skill in the OpenClaw ecosystem. Supports PR creation, code review, issue management, and branch operations — complete Git workflows through conversation.

clawhub install openclaw/github

Typical usage:

"Review the latest PR on my repo and summarize the changes"
"Create a new issue for the login bug with reproduction steps"

Security-check (#12, Score 60/80)

Automatically scans project dependencies and code for security vulnerabilities, integrated with CVE databases.

clawhub install community/security-check

Linear (#14, Score 59/80)

Integrates with Linear project management — create/update issues and check sprint progress through conversation.

clawhub install community/linear-claw

Cron-backup (#20, Score 57/80)

Scheduled backups of OpenClaw configuration and memory data to prevent data loss.

clawhub install community/cron-backup

🔍 Search & Research

Tavily (#4, Score 67/80)

A search engine API designed specifically for AI Agents. Returns structured results — faster and more precise than Web Browsing. Requires a Tavily API Key.

clawhub install framix-team/openclaw-tavily

Web Browsing (#2, Built-in)

OpenClaw’s foundational capability — lets the Agent browse pages, extract content, and follow links. Most workflows depend on this.

Felo Search (#15, Score 59/80)

Multi-language search engine integration with strong CJK language support.

clawhub install community/felo-search

📋 Productivity

GOG (#3, Score 68/80)

“Get Organized with GOG” — task planning and goal tracking. Helps the Agent break complex tasks into executable steps. Top 3 in download count.

clawhub install openclaw/gog

Obsidian (#9, Score 62/80)

Integrates with Obsidian notes — the Agent can read/write your knowledge base, create notes, and search existing content.

clawhub install community/obsidian-claw

Notion (#13, Score 59/80)

Read/write operations on Notion databases and pages. Great for teams using Notion for project management.

clawhub install community/notion-claw

Summarize (#19, Score 58/80)

Long document/webpage summarization with a high security score (9/10). Ideal for processing large amounts of reading material.

clawhub install community/summarize

🤖 Automation

n8n (#8, Score 63/80)

Connects OpenClaw with the n8n workflow engine for cross-platform automation. Example: receive email → analyze content → create Jira issue → notify Slack.

clawhub install community/n8n-openclaw

Browser Automation (#16, Score 58/80)

Playwright-based browser automation — fill forms, click buttons, take screenshots, handle sites requiring login.

clawhub install community/browser-automation

Home Assistant (#10, Score 61/80)

Control smart home devices through conversation. “Turn off the living room lights.” “Set the AC to 24°C.”

clawhub install openclaw/homeassistant

🧠 AI/ML Advanced

Capability Evolver (#11, Score 60/80)

Enables the Agent to self-improve — analyzes failed tasks and automatically adjusts strategies. Extremely high learning value (LRN 9/10).

clawhub install community/capability-evolver

RAG Pipeline (#38, Score 51/80)

Adds Retrieval-Augmented Generation to OpenClaw — build vector indexes on local documents and retrieve relevant context.

clawhub install community/rag-pipeline

📊 Data Processing

Firecrawl (#18, Score 58/80)

Structured web scraping with pagination support, JavaScript rendering, and clean Markdown/JSON output.

clawhub install community/firecrawl

DuckDB CRM (#28, Score 55/80)

Lightweight local CRM powered by DuckDB. Manage contacts and customer data with SQL — all data stays local.

clawhub install community/duckdb-crm

Software Engineer

clawhub install openclaw/github
clawhub install community/security-check
clawhub install community/n8n-openclaw

Researcher / Knowledge Worker

clawhub install framix-team/openclaw-tavily
clawhub install community/obsidian-claw
clawhub install community/summarize

Project Manager

clawhub install openclaw/gog
clawhub install community/notion-claw
clawhub install community/todoist-claw

Content Creator

clawhub install community/image-generation
clawhub install community/felo-slides
clawhub install community/youtube-digest

Security Reminders

  1. Only install Skills you need — each Skill has access to your data
  2. Prefer official Skills (openclaw/ prefix) and high-star community Skills
  3. Review source code before installing — pay attention to network requests and file access
  4. Use Skill Vetter — a community-built security scanner:
    clawhub install community/skill-vetter
    openclaw run "Scan the skill community/xxx for security issues"
  5. Configure least privilege — restrict allowed_paths and denied_paths in ~/.openclaw/config.yaml

🔒 Security Auditing

Skill Vetter (Rating 4.9/5, 695 Stars)

Install this before installing anything else. A security-first vetting protocol that enforces a standardized checklist before any third-party Skill installation. It’s a docs-only Skill — zero executable code, zero permissions required.

clawhub install openclaw/skill-vetter

The vetting protocol has 4 steps:

  1. Source Check — Author reputation, star count, last update date
  2. Code Review (MANDATORY) — File-by-file inspection for red flags:
    • curl/wget to unknown URLs
    • Reading ~/.ssh, ~/.aws, ~/.config
    • eval()/exec() with external input
    • base64 decoding, obfuscated code, sudo requests
    • Accessing browser cookies/sessions
  3. Permission Scope Analysis — Are file reads/writes, commands, and network access minimal?
  4. Risk Classification — Four-tier system:
Risk Level Examples Action
🟢 Low Notes, weather, formatting Basic review, install OK
🟡 Medium File ops, browser, APIs Full code review required
🔴 High Credentials, trading, system Human approval required
⛔ Extreme Security configs, root access Do NOT install

Produces a standardized report with red flag detection results, permission inventory, and final verdict (✅ Safe / ⚠️ Caution / ❌ Do Not Install).

Trust hierarchy: Official Skills > High-star repos (1000+) > Known authors > Unknown sources > Skills requesting credentials (always require human approval)


🧬 Self-Evolution

Self-Improvement (v2, 2.6k Downloads)

Enables the Agent to learn from mistakes and continuously improve. It captures errors, user corrections, feature requests, and learnings into structured Markdown files for future sessions to reference.

clawdhub install self-improving-agent

How it works:

  • Automatic detection triggers — Logs automatically when commands fail, users correct output, APIs break, or knowledge is outdated
  • Structured logging — Categorized storage in the .learnings/ directory:
    .learnings/
    ├── LEARNINGS.md          # Corrections, knowledge gaps, best practices
    ├── ERRORS.md             # Command failures, exceptions, timeouts
    └── FEATURE_REQUESTS.md   # User-requested capabilities
  • Learning promotion — When a learning recurs (≥3 times, across 2+ tasks, within 30 days), it’s automatically promoted to project-level memory files:
    • SOUL.md — Behavioral guidelines
    • AGENTS.md — Workflow improvements
    • TOOLS.md — Tool usage gotchas
  • Cross-session communication — Share learnings between sessions via sessions_send
  • Automatic Skill extraction — When a learning is generic enough, it can be extracted into a standalone reusable Skill

Typical scenario:

Agent first time: runs npm install → fails
Agent logs: project uses pnpm, not npm
Agent second time: runs pnpm install → succeeds

Supports OpenClaw, Claude Code, Codex CLI, GitHub Copilot, and other Agent platforms.


For more Skills, check the ClawHub Marketplace and OpenClaw Top 50 Skills.