ClawHub

ClawHub is OpenClaw's community skill marketplace — an open-source registry (8.9k GitHub stars) where developers publish, version, and share agent skills. With over 10,700 skills as of mid-2026, it's the largest source of pre-built agent capabilities.

:::warning Security first ClawHub has been targeted by malicious actors — the ClawHavoc campaign found hundreds of malicious skills in early 2026. Always verify skills before installing. See the Security section and Skill Verification for the full checklist. :::

How Skills Work

A skill is a directory containing a SKILL.md file — YAML frontmatter for metadata plus Markdown instructions that tell the agent what to do when the skill activates.

skills/git-summary/SKILL.md
---
name: git-summary
description: Summarize recent git activity
trigger: "git summary|what changed|recent commits"
tools:
  - shell
---

## Git Summary

When triggered, analyze the recent git history:
1. Run `git log --oneline -20` to get recent commits
2. Run `git diff --stat HEAD~5` to see what changed
3. Summarize the activity: who committed, what changed, any patterns
4. Present a clear, concise summary

Frontmatter Fields

Field	Required	Description
`name`	Yes	Unique slug (`^[a-z0-9][a-z0-9-]*$`)
`description`	Yes	Brief description of what the skill does
`version`	No	Semver (auto-incremented on publish)
`trigger`	No	Activation pattern — keywords, regex, or `"*"` for always-on
`tools`	No	Required tools: `shell`, `filesystem`, `browser`, `http`, `memory`, `chat`
`author`	No	Skill author name
`config`	No	User-configurable settings
`depends`	No	Other skills this skill requires

Triggers

# Keyword trigger — activates on matching phrases
trigger: "deploy|ship it|push to prod"

# Regex trigger
trigger: "/^remind me .+/i"

# Always-on — loaded into every message (expensive)
trigger: "*"

# Attachment trigger — fires on media type
trigger: "attachment:image"
trigger: "attachment:audio"

Tools

Tool	Capability
`shell`	Execute shell commands
`filesystem`	Read/write files
`browser`	Navigate web pages, take screenshots
`http`	Make HTTP requests to APIs
`memory`	Read/write persistent memory
`chat`	Send messages to channels

Conditional Activation

Skills can declare runtime requirements so they only activate when dependencies are available:

---
name: docker-manager
description: Manage Docker containers
tools:
  - shell
metadata:
  openclaw:
    requires:
      bins:
        - docker
        - docker-compose
      env:
        - DOCKER_HOST
      platform:
        - linux
        - darwin
---

If docker isn't on PATH or the OS doesn't match, the skill is silently skipped.

Multi-File Skills

Skills can include support files alongside SKILL.md:

skills/code-reviewer/
├── SKILL.md           # Main skill definition
├── templates/
│   └── review.md      # Review template
├── scripts/
│   └── lint.sh        # Helper script
└── examples/
    └── sample.md      # Usage example

Bundle size is capped at 50 MB of text-only files.

Browsing & Discovery

CLI Search

# Search by keyword
openclaw clawhub search "email"
openclaw clawhub search "github automation"
openclaw clawhub search "home assistant"

# Browse by category
openclaw clawhub browse --category productivity
openclaw clawhub browse --category development
openclaw clawhub browse --category home-automation
openclaw clawhub browse --category research
openclaw clawhub browse --category creative

# View full details
openclaw clawhub info email-triage

Web Interface

Browse visually at openclaw.ai/clawhub — search, filter by category, read source code, and check security reports.

Skill Categories

Category	Examples
Development	Code review, git automation, PR management, testing, CI/CD
Productivity	Email triage, calendar management, note-taking, task tracking
Research	Web search, news monitoring, data analysis, summarization
Home Automation	Home Assistant, IoT control, smart home routines
Creative	Image generation, writing assistance, content creation
Communication	Message drafting, translation, social media
DevOps & Cloud	Docker, Kubernetes, AWS, monitoring, deployment
Security	Vulnerability scanning, audit logging, threat detection
Finance	Budget tracking, crypto monitoring, invoice processing
Media	Image/video generation, audio transcription, editing

Curated Lists

Beyond ClawHub's built-in discovery:

Awesome OpenClaw Skills — 2,999 curated skills across 20+ categories, community-maintained
skills.sh — Alternative skill browser with additional filtering
OpenClaw Skills Archive — Version-pinned snapshots of ClawHub skills

Recommended Starter Skills

For new users, these are well-established and widely used:

Skill	What It Does	Category
`git-summary`	Summarize recent git activity	Development
`email-triage`	Prioritize and summarize inbox	Productivity
`daily-standup`	Generate daily status reports	Productivity
`web-search`	Search the web and summarize results	Research
`code-reviewer`	Review code changes with feedback	Development
`meeting-notes`	Extract decisions and action items	Productivity
`clawsec`	Security scanning and SOUL.md monitoring	Security
`clawdex`	Pre-installation malicious skill scanning	Security

Installing Skills

From ClawHub

# Install latest version
openclaw clawhub install email-triage

# Install a specific version
openclaw clawhub install email-triage@2.1.0

# Install with sandbox restrictions
openclaw clawhub install email-triage --sandbox --no-shell

From a URL

# Install from any Git repository
openclaw clawhub install https://github.com/user/skill-repo/skill.md

Verify Before Installing

Always check security before installing:

# Check security scan results
openclaw clawhub security-report email-triage

# Scan with Clawdex (malicious skills database)
openclaw clawhub scan email-triage

# Review the source code
openclaw clawhub view email-triage

# Check author and community signals
openclaw clawhub info email-triage

See Skill Verification for the complete security checklist.

After installation, the skill is available immediately — no gateway restart needed.

Managing Skills

# List all installed skills
openclaw skill list

# Check for updates
openclaw clawhub outdated

# Update all skills
openclaw clawhub update

# Update a specific skill
openclaw clawhub update email-triage

# Disable a skill (keep installed but inactive)
openclaw skill disable email-triage

# Re-enable
openclaw skill enable email-triage

# Remove completely
openclaw skill remove email-triage

# Test a skill without installing
openclaw skill test ./my-skill.md "test trigger message"

Configuring Skills

Some skills expose configuration options:

# View current config
openclaw skill config email-triage

# Set a value
openclaw skill config email-triage --set priority_threshold=high
openclaw skill config email-triage --set summary_length=brief

Version Pinning

For production stability, pin skill versions:

# Install a specific version
openclaw clawhub install email-triage@2.1.0

# Don't auto-update this skill
openclaw skill config email-triage --set auto_update=false

Publishing Skills

Share your skills with the community. All ClawHub skills are licensed under MIT-0 (no attribution required, no paid skills, no per-skill license overrides).

Three Publishing Pathways

Pathway	Best For	How
Local creation	Personal use, testing	Create in `~/.openclaw/workspace/skills/`
Skill Workshop	Governed environments	Proposal system with approval workflow
ClawHub publish	Community sharing	Public registry via CLI

Publishing to ClawHub

# Login with GitHub account
openclaw clawhub login

# Validate your skill (checks format, required fields)
openclaw clawhub validate ./my-skill/SKILL.md

# Publish
openclaw clawhub publish ./my-skill/SKILL.md

# Publish an update (bumps version)
openclaw clawhub publish ./my-skill/SKILL.md --update

Publishing Requirements

Valid YAML frontmatter with name and description
Name must be unique on ClawHub (slug format: ^[a-z0-9][a-z0-9-]*$)
Must pass automated security scanning (VirusTotal + code analysis)
Must include at least one usage example in the Markdown body
GitHub account must be at least 14 days old (raised from one week; confirmed by the project creator)
Bundle under 50 MB, text-only files

What Happens After Publishing

Validation — Token, metadata, name, version, and files are checked
Staged publish (July 2026+) — Publishes are held behind a prepublication worker and scanned before finalization
Security scan — VirusTotal + code pattern analysis + ClawScan runs automatically
Hold period — New releases may be held from install surfaces until review completes; scan-held/blocked releases stay visible to their owners but are hidden from public catalogs
Available — Once cleared, the skill appears in search and can be installed

If validation fails, nothing is published and you get an error explaining why.

Skill Workshop (Governed Publishing)

For teams or environments that need approval before skills go live:

# Propose a new skill (goes to pending review)
openclaw skill-workshop propose-create ./my-skill/SKILL.md

# List pending proposals
openclaw skill-workshop list

# Inspect a proposal
openclaw skill-workshop inspect proposal-id

# Apply (approve) a proposal
openclaw skill-workshop apply proposal-id

# Reject a proposal
openclaw skill-workshop reject proposal-id --reason "needs security review"

The Workshop supports hash binding, rollback, and scanner gating (OpenClaw static analysis + VirusTotal + NVIDIA SkillSpector). See the Skill Workshop guide for full details.

Security

The ClawHavoc Incident

In early February 2026, three independent security firms discovered widespread malicious skills on ClawHub:

Firm	Finding	Method
Koi Security	341 malicious skills out of 2,857 total	Full audit of every skill
SlowMist	472 malicious skills sharing C2 infrastructure	Threat intelligence tracking
Snyk	76 confirmed malicious payloads	Human-in-the-loop code review

The malicious skills deployed multiple attack types:

AMOS (Atomic macOS Stealer) — targeting 60+ cryptocurrency wallets
Windows infostealers with keylogging
Reverse shells connecting to command-and-control servers
Credential exfiltration — .env files sent to attacker-controlled webhooks
Two-stage delivery — Base64-encoded payloads, curl-then-bash loading
Fake system dialogs — password phishing via OS-native prompts

100% of confirmed malicious skills contained malicious code, while 91% also employed prompt injection techniques.

Unit 42 Findings (June 2026)

On June 23, 2026, Palo Alto Networks' Unit 42 published research identifying five malicious skills that evaded ClawHub's scanning during their February–May 2026 analysis:

Two macOS infostealers with command-and-control connectivity (linked to cluw and Atomic macOS Stealer)
omnicogg — a scanner-evasion skill using 22 MB of file padding to exceed content-analysis size limits
money-radar — runtime affiliate-link injection
letssendit — an agentic Solana front-running skill

OpenClaw banned the publisher accounts and deleted all five skills. The takeaway: scanning keeps improving (see below), but evasion keeps pace — always review skill source before installing.

Broader Quality Issues

Snyk's ToxicSkills audit found problems beyond just intentionally malicious skills:

36.8% of skills have at least one security flaw
13.4% contain critical-level issues
7.1% leak credentials (283 skills)

Current Security Measures

Automated scanning:

VirusTotal integration (v2026.2.6+) — All skills scanned on upload, periodic re-scanning
NVIDIA SkillSpector (June 1, 2026 partnership) — Static checks plus AI-assisted semantic analysis in the ClawScan pipeline, producing Clean/Suspicious/Malicious verdicts, along with Skill Cards for provenance
Staged publishing (July 2026+) — Publishes held behind prepublication checks before finalization
Code pattern analysis — Checks declared frontmatter vs actual behavior, flags mismatches
Metadata validation — Detects undeclared environment variables and binary dependencies

Moderation: Signed-in users can report skills; a skill is auto-hidden after 4 unique abuse reports, and moderators can hide, restore, or ban.

Community tools:

Clawdex (7.1k downloads) — Pre-installation scanning against Koi's malicious skills database
SkillGuard — Skill file vulnerability scanner
SafeClaw Scanner — Third-party malicious pattern detection

Known Limitations

Trail of Bits demonstrated in June 2026 that current scanners can be bypassed, concluding that no LLM-based scanning can reliably detect all malicious skills. This means:

Automated scanning is necessary but not sufficient
Always review skill source code before installing
Use sandboxing (--sandbox, --no-shell) for untrusted skills
Prefer well-known skills with high download counts and active maintenance
Install Clawdex for an additional layer of pre-install scanning

Security Checklist (Quick Reference)

# Before installing ANY skill:

# 1. Check the security report
openclaw clawhub security-report <skill-name>

# 2. Scan with Clawdex
openclaw clawhub scan <skill-name>

# 3. Review the source code
openclaw clawhub view <skill-name>

# 4. Check author reputation and download count
openclaw clawhub info <skill-name>

# 5. Install with restrictions if unsure
openclaw clawhub install <skill-name> --sandbox --no-shell

Red Flags

Avoid skills that contain:

curl | bash or wget | bash patterns
Base64-encoded strings (obfuscation)
Zero-width Unicode characters (hidden content)
Hardcoded URLs to unknown domains
Requests to access ~/.ssh, ~/.aws, or credential paths
chmod 777 or permission loosening
SOUL.md modification instructions

Disabling ClawHub

For maximum security, disable ClawHub entirely:

~/.openclaw/openclaw.json
{
  "skills": {
    "allow_install": false,
    "allow_clawhub": false
  }
}

Reporting Malicious Skills

openclaw clawhub report <skill-name> --reason malicious

Reports are reviewed by the ClawHub team and forwarded to VirusTotal.

Building Your Own Skills

Minimal Example

skills/weather/SKILL.md
---
name: weather-check
description: Check the weather for a location
trigger: "weather|forecast|temperature"
tools:
  - http
---

## Weather Check

When the user asks about weather:
1. Determine the location from their message
2. Call a weather API (e.g., wttr.in) to get current conditions
3. Present temperature, conditions, and forecast in a brief summary

With Configuration

skills/daily-digest/SKILL.md
---
name: daily-digest
description: Generate a daily summary of news and tasks
trigger: "daily digest|morning brief|what's happening"
tools:
  - http
  - memory
config:
  news_sources:
    type: string
    default: "hackernews,techcrunch"
  include_calendar:
    type: boolean
    default: true
---

## Daily Digest

Generate a personalized morning briefing:
1. Fetch top stories from configured news sources
2. If calendar integration is enabled, check today's schedule
3. Review pending tasks from memory
4. Present a concise briefing organized by priority

With Dependencies

---
name: full-code-review
description: Complete code review pipeline
depends:
  - git-summary
  - code-reviewer
tools:
  - shell
  - filesystem
---

Testing

# Test locally without installing
openclaw skill test ./skills/weather/SKILL.md "What's the weather in London?"

# Test with a specific model
openclaw skill test ./skills/weather/SKILL.md "forecast for Tokyo" --model claude-haiku-4-5-20251001

See the Skill Development guide for advanced patterns including state management, dependency chaining, and multi-step workflows.

Customizing Installed Skills

You can modify installed skills to fit your needs:

# Find where skills are stored
ls ~/.openclaw/skills/

# Edit an installed skill
nano ~/.openclaw/skills/email-triage/SKILL.md

Common customizations:

Adjust triggers — change activation phrases to match your vocabulary
Modify instructions — tailor the agent's behavior for your workflow
Change tools — remove shell access if not needed for security
Add config — expose settings you want to tweak

Changes take effect immediately — no restart needed. Note that openclaw clawhub update will overwrite your changes, so disable auto-update for customized skills.

CLI Reference

# Discovery
openclaw clawhub search <query>
openclaw clawhub browse --category <category>
openclaw clawhub info <skill-name>
openclaw clawhub view <skill-name>            # View source code

# Installation
openclaw clawhub install <skill-name>[@version]
openclaw clawhub install <url>
openclaw clawhub install <skill-name> --sandbox --no-shell

# Management
openclaw skill list
openclaw skill disable <name>
openclaw skill enable <name>
openclaw skill remove <name>
openclaw skill config <name> [--set key=value]
openclaw skill test <path> "<trigger message>"

# Updates
openclaw clawhub outdated
openclaw clawhub update [skill-name]

# Security
openclaw clawhub security-report <skill-name>
openclaw clawhub scan <skill-name>
openclaw clawhub validate ./skill.md
openclaw security scan [--all]

# Publishing
openclaw clawhub login
openclaw clawhub publish ./skill.md [--update]
openclaw clawhub report <skill-name> --reason <reason>

How Skills Work​

Frontmatter Fields​

Triggers​

Tools​

Conditional Activation​

Multi-File Skills​

Browsing & Discovery​

CLI Search​

Web Interface​

Skill Categories​

Curated Lists​

Recommended Starter Skills​

Installing Skills​

From ClawHub​

From a URL​

Verify Before Installing​

Managing Skills​

Configuring Skills​

Version Pinning​

Publishing Skills​

Three Publishing Pathways​

Publishing to ClawHub​

Publishing Requirements​

What Happens After Publishing​

Skill Workshop (Governed Publishing)​

Security​

The ClawHavoc Incident​

Unit 42 Findings (June 2026)​

Broader Quality Issues​

Current Security Measures​

Known Limitations​

Security Checklist (Quick Reference)​

Red Flags​

Disabling ClawHub​

Reporting Malicious Skills​

Building Your Own Skills​

Minimal Example​

With Configuration​

With Dependencies​

Testing​

Customizing Installed Skills​

CLI Reference​

See Also​