Skip to main content

Module 4: ClawHub Marketplace

Learning Objectives​

By the end of this module, you will be able to:

  • Use the clawhub CLI to search, install, and manage Skills
  • Understand ClawHub's review mechanisms and trust model
  • Publish your own Skills to ClawHub
  • Explain the ClawHavoc incident and the resulting security enhancements
  • Evaluate Skill security and identify potential risks
Prerequisites

Please complete Module 3: Skills System & SKILL.md Specification first, ensuring you understand Skill structure and development.

ClawHub Overview​

ClawHub is the official OpenClaw Skill marketplace, similar to npm, PyPI, or Docker Hub, but purpose-built for AI Agent Skills. As of March 2026, ClawHub hosts over 13,000 Skills spanning productivity, developer tools, data analysis, automation, and more.

ClawHub Ecosystem​

Developers                ClawHub                     Users
┌──────────┐           ┌──────────────────┐        ┌──────────┐
│ Develop  │           │                  │        │ Search   │
│ Skill ──▶│── push ──▶│  Skill Registry  │◀─ search─│ ────────▶│
│          │           │  ┌────────────┐  │        │          │
│ SKILL.md │           │  │ 13,000+    │  │        │ Install  │
│ Code     │           │  │  Skills    │  │        │ ────────▶│
│ Tests    │           │  └────────────┘  │        │          │
└──────────┘           │                  │        │ Rate     │
                       │  ┌────────────┐  │        │ ────────▶│
                       │  │ Review     │  │        └──────────┘
                       │  │ Pipeline   │  │
                       │  │ VirusTotal │  │
                       │  │ Static     │  │
                       │  │ Analysis   │  │
                       │  │ Signature  │  │
                       │  └────────────┘  │
                       └──────────────────┘

Key Metrics​

MetricValue
Total Skills13,000+
Verified Developers2,800+
Official Skills150+
Daily Installs45,000+
Average Review Time24 hours
VirusTotal Scan Rate100% (mandatory)

Using the ClawHub CLI​

Searching for Skills​

# Basic search
clawhub search weather

# Filter by tag
clawhub search --tag productivity

# Filter by author
clawhub search --author openclaw-official

# Show only verified Skills
clawhub search weather --verified

# Sort by download count
clawhub search --sort downloads --limit 20

Example search results:

NAME                          AUTHOR              DOWNLOADS  RATING  VERIFIED
─────────────────────────────────────────────────────────────────────────────
weather-lookup                openclaw-official    124,500    4.9     ✓
weather-forecast-pro          weather-dev          45,200     4.7     ✓
weather-alerts                community-tools      12,300     4.5     ✓
weather-radar                 wx-enthusiast         8,700     4.3
weather-historical            data-archive          3,200     4.1

Installing Skills​

# Standard install
clawhub install openclaw-official/weather-lookup

# Install a specific version
clawhub install openclaw-official/weather-lookup@1.2.0

# Install without interactive confirmation (CI/CD scenarios)
clawhub install openclaw-official/weather-lookup --yes

# View the security report before installing
clawhub inspect openclaw-official/weather-lookup

Installation flow:

clawhub install openclaw-official/weather-lookup
╭─────────────────────────────────────────────╮
│  Installing weather-lookup v1.2.0           │
│  Author: openclaw-official (verified ✓)     │
│                                             │
│  Permissions requested:                     │
│    ✓ network:api.openweathermap.org         │
│    ✓ network:api.weatherapi.com             │
│                                             │
│  Security scan:                             │
│    ✓ VirusTotal: 0/72 detections            │
│    ✓ Static analysis: passed                │
│    ✓ Signature: valid (SHA-256)             │
│                                             │
│  Accept? [Y/n]                              │
╰─────────────────────────────────────────────╯

Managing Installed Skills​

# List all installed Skills
clawhub list

# View details for a specific Skill
clawhub info weather-lookup

# Update a single Skill
clawhub update weather-lookup

# Update all Skills
clawhub update --all

# Remove a Skill
clawhub remove weather-lookup

# View a Skill's changelog
clawhub changelog weather-lookup

ClawHub Trust Model​

ClawHub uses a multi-tiered trust model to help users evaluate Skill reliability:

Trust Levels​

LevelBadgeDescriptionReview Process
Official🏛Developed and maintained by the OpenClaw teamInternal review
Verified✓Verified developer, passed full reviewAutomated + manual review
Community--Community contribution, passed basic scanningAutomated scanning
Unverified⚠️Newly published, review not yet completeQueued
Flagged:triangular_flag:Reported by the community as problematicUnder investigation
Security Advice
  • Prefer installing Official or Verified Skills
  • Before installing a Community Skill, always check its permission declarations
  • Never install a Flagged Skill
  • Use clawhub inspect to view the full security report before installation

Security Scanning Pipeline​

Every Skill uploaded to ClawHub goes through the following scans:

Upload → SHA-256 signature verification → VirusTotal scan (72 engines) →
Static code analysis → Permission audit → Dependency check → Sandbox test run →
[Manual review (Verified and above)] → Publish
  1. Signature verification: Confirms file integrity and that it hasn't been tampered with
  2. VirusTotal scan: Scans with 72 antivirus engines
  3. Static analysis: Checks for suspicious code patterns (e.g., eval(), obfuscated code, suspicious network requests)
  4. Permission audit: Confirms declared permissions match actual code behavior
  5. Dependency check: Scans npm/pip dependencies for known vulnerabilities
  6. Sandbox test: Runs in an isolated environment to verify expected behavior

ClawHavoc Incident Retrospective​

Major Security Incident

ClawHavoc (January 2026) was the most serious supply chain attack in OpenClaw's history. Attackers uploaded 2,400 malicious Skills to ClawHub disguised as legitimate tools, which secretly exfiltrated users' API keys, environment variables, and files.

Event Timeline​

DateEvent
2026-01-08Attackers begin mass-uploading malicious Skills
2026-01-15Community members report anomalous behavior
2026-01-16OpenClaw team confirms the attack and begins investigation
2026-01-17Emergency takedown of 2,400 malicious Skills
2026-01-20Security update released, mandatory VirusTotal scanning enforced
2026-01-25New trust model and review pipeline rolled out
2026-02-01All existing Skills complete re-scanning

Improvements After ClawHavoc​

  1. Mandatory VirusTotal scanning: All Skills must pass 72-engine scanning before upload
  2. Permission transparency: All requested permissions are clearly displayed during installation
  3. Enhanced static analysis: New suspicious pattern detection (obfuscated code, hidden network requests)
  4. Developer identity verification: Verified Developer program introduced
  5. Community reporting mechanism: One-click reporting for suspicious Skills
  6. Automatic revocation: Skills with VirusTotal detection rate > 0 are automatically delisted

Self-Check​

If you installed any Skills during the ClawHavoc period, run the following immediately:

# Check for known malicious Skills
clawhub audit

# Example output:
# Auditing installed skills...
# ✓ 45/47 skills passed security check
# ⚠ 2 skills require attention:
#   - suspicious-tool v1.0.0 (REMOVED from ClawHub)
#   - fake-helper v0.3.2 (FLAGGED: potential data exfiltration)
#
# Run 'clawhub remove <skill>' to uninstall flagged skills
# Run 'clawhub audit --rotate-keys' to rotate compromised API keys

# Remove suspicious Skills and rotate API Keys
clawhub remove suspicious-tool fake-helper
clawhub audit --rotate-keys

Hands-On: Publishing Your Skill to ClawHub​

Let's publish the Pomodoro Timer Skill created in Module 3 to ClawHub.

Step 1: Create a ClawHub Account​

# Register an account
clawhub register

# Or log in to an existing account
clawhub login

Step 2: Prepare for Publishing​

cd ~/.openclaw/skills/local/pomodoro-timer

# Validate the Skill structure
clawhub validate .

# Output:
# Validating pomodoro-timer...
# ✓ SKILL.md: valid
# ✓ index.js: found
# ✓ Permissions: none requested (safe)
# ✓ Runtime: node:20-slim (supported)
# ✓ Version: 0.1.0 (valid semver)
# Ready to publish!

Step 3: Write a README.md (for the ClawHub page)​

cat > README.md << 'EOF'
# Pomodoro Timer Skill

A simple yet effective Pomodoro timer for OpenClaw.

## Features

- Start, pause, and reset Pomodoro sessions
- Customizable work and break durations
- Session tracking and statistics
- Multi-language support (zh-TW, en, ja)

## Usage

Just tell your OpenClaw agent:

- "Start a pomodoro"
- "Pomodoro status"
- "Pause the pomodoro"
- "Reset the pomodoro"

## Installation

```bash
clawhub install your-username/pomodoro-timer

License​

MIT EOF


### Step 4: Publish

```bash
# Publish to ClawHub
clawhub publish .

# Output:
# Publishing pomodoro-timer v0.1.0...
# ✓ Package created (12.3 KB)
# ✓ SHA-256 signature generated
# ✓ Uploaded to ClawHub
# ✓ VirusTotal scan queued
#
# Your skill is now available at:
# https://clawhub.dev/your-username/pomodoro-timer
#
# Note: Full security scan takes ~24 hours.
# Status: Unverified → Community (after scan passes)

Step 5: Manage Published Skills​

# View all your published Skills
clawhub my-skills

# View download statistics
clawhub stats pomodoro-timer

# Publish an updated version
# 1. Update the version in SKILL.md
# 2. Re-publish
clawhub publish . --bump patch  # 0.1.0 → 0.1.1

# Unpublish a specific version
clawhub unpublish pomodoro-timer@0.1.0

# Transfer ownership
clawhub transfer pomodoro-timer --to new-owner

Skill Ratings & Feedback​

Rating System​

ClawHub uses a five-star rating system with text reviews:

# Rate a Skill
clawhub rate weather-lookup --stars 5 --comment "Works great, fast responses!"

# View reviews
clawhub reviews weather-lookup

# Report an issue
clawhub report weather-lookup --reason "suspicious-behavior" \
  --details "Skill attempts to access undeclared network endpoints"

Skill Selection Checklist​

Before installing a community Skill, evaluate it using this checklist:

  • Is the trust level Verified or higher?
  • Are the permission declarations reasonable? (A search Skill shouldn't need filesystem:write)
  • Is the rating 4.0 or above?
  • Has the last update been within 6 months?
  • Does it have sufficient downloads and reviews?
  • Is the source code open for review?
  • Does clawhub inspect show a clean report?

Common Errors & Troubleshooting​

Error 1: Publish Failed -- Name Already Taken​

Error: Skill name 'weather-lookup' is already taken

Fix: Choose a different name, or add your namespace prefix (e.g., your-username/my-weather-lookup).

Error 2: VirusTotal Scan Failed​

Error: VirusTotal detected 2/72 positives

Fix:

# View the detailed report
clawhub scan-report pomodoro-timer

# If it's a false positive, file an appeal
clawhub appeal pomodoro-timer --reason "False positive: detected pattern is standard API call"

Error 3: Install Failed -- Version Conflict​

Error: Dependency conflict: skill-a requires node:18 but skill-b requires node:20

Fix: Each Skill runs in an independent container, so runtime conflicts typically don't occur. If you encounter one, update to the latest version.

Error 4: Permission Audit Failed​

Error: Permission audit failed: code accesses 'api.hidden-endpoint.com' but only declares 'api.example.com'

Fix: Correctly declare all network access targets in SKILL.md.

Exercises​

  1. Skill Discovery: Use clawhub search to find the top 10 Skills by downloads. Analyze their common characteristics (permission requirements, feature types, author distribution).

  2. Security Audit: Run clawhub audit on all your installed Skills, record the results, and address any warnings.

  3. Publishing Practice: Publish the Pomodoro Timer you developed in Module 3 to ClawHub (a test environment is fine), completing the full develop-to-publish workflow.

  4. Evaluation Exercise: Select 5 Community-level Skills. Use this module's checklist to evaluate their security and record your findings.

Quiz​

  1. How many Skills are currently on ClawHub?

    • A) 5,000+
    • B) 8,000+
    • C) 13,000+
    • D) 20,000+

  2. What is the correct command to install a Skill?

    • A) openclaw install weather-lookup
    • B) clawhub install openclaw-official/weather-lookup
    • C) npm install weather-lookup
    • D) pip install weather-lookup

  3. How many malicious Skills were uploaded during the ClawHavoc incident?

    • A) 100
    • B) 500
    • C) 2,400
    • D) 10,000

  4. Which trust level indicates a Skill has passed full review?

    • A) Community
    • B) Unverified
    • C) Verified
    • D) Flagged

  5. Which security measure became mandatory after ClawHavoc?

    • A) Manual review
    • B) Two-factor authentication
    • C) VirusTotal scanning
    • D) Blockchain signatures
View Answers
  1. C -- As of March 2026, ClawHub hosts over 13,000 Skills.
  2. B -- Use the clawhub install <author>/<skill> format to install Skills.
  3. C -- 2,400 malicious Skills were uploaded during the ClawHavoc incident.
  4. C -- Verified indicates the Skill was published by a verified developer and has passed both automated and manual review.
  5. C -- After ClawHavoc, VirusTotal scanning (72 engines) became a mandatory requirement for all Skill uploads.

Next Steps​

You now have a solid grasp of the ClawHub marketplace and its security concepts. Next, let's explore OpenClaw's memory system and learn how the Agent maintains context across conversations.

Go to Module 5: Persistent Memory & Personalization →