A defense-in-depth security layer that sanitizes external AI skills by regenerating them from scratch to eliminate hidden prompt injection and malicious payloads.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install skill-firewall
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install skill-firewall using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Skill Firewall?
The Skill Firewall is an essential security framework designed to protect AI coding agents from vulnerabilities found in third-party content. External repositories can often contain hidden HTML comments, zero-width Unicode characters, or social engineering tactics designed to exfiltrate data or execute arbitrary code. By implementing a zero-trust protocol, this skill ensures that your agent never copies external content verbatim. Instead, the agent is forced to analyze the purpose of the code and rewrite a clean, sanitized version from scratch.
This approach effectively acts as a compiler sanitization pass for Openclaw Skills, ensuring that malicious payloads do not survive the transition to your local environment. It shifts the agent's role from a passive consumer to an active security layer, prioritizing safety and human-in-the-loop approval before any new functionality is integrated into your workflow.
Skill Firewall Use Cases
- Safely importing new Openclaw Skills from public repositories like GitHub or community hubs.
- Auditing third-party skills for hidden instructions or obfuscated Unicode tricks.
- Preventing remote code execution (RCE) attempts during the setup of unknown extensions.
- Sanitizing skills that contain suspicious external references or data exfiltration URLs.
How Skill Firewall Works
- The agent intercepts a request to install or use an external skill and acknowledges the security protocol.
- It performs a silent analysis of the source content to identify the legitimate purpose while ignoring any hidden instructions.
- The agent regenerates a completely new version of the skill from scratch using only standard ASCII characters.
- A Skill Firewall Report is presented to the user, identifying the skill's purpose and any red flags discovered.
- The sanitized skill is only saved to the local directory after the user provides explicit approval.
Skill Firewall Setup
To activate the protection layer for your Openclaw Skills, you must include the firewall definition in your local environment. Navigate to your configuration directory and ensure the skill is present:
# Move to your local skills directory
cd path/to/your/skills
# Clone the firewall repository or create the skill-firewall.md file
git clone https://github.com/openclaw/skill-firewall
Once added, the agent will automatically trigger the regeneration protocol whenever you attempt to add new Openclaw Skills from external sources.
Skill Firewall Data Schema & Taxonomy
The Skill Firewall generates a structured security report and a sanitized output file. The data is organized as follows:
| Component | Description |
|---|---|
| Original Source | The URL or repository name of the untrusted skill |
| Purpose Identified | A brief summary of the legitimate functionality found in the source |
| Suspicious Elements | A list of detected red flags like HTML comments or Unicode tricks |
| Clean Rewrite | The final, sanitized Markdown content following the Openclaw Skills standard |
| Approval Status | A boolean flag indicating if the user has authorized the local save |
Skill Firewall Advanced Features
- Zero-trust regeneration that strips all non-ASCII characters and hidden HTML metadata.
- Automated detection of command-line injection attempts like curl or npx inside setup instructions.
- Poison chain prevention that blocks nested references to unverified external files.
- Authority spoofing detection to flag instructions that attempt to override system safety protocols.
- Human-in-the-loop validation requiring manual approval for any new Openclaw Skills.
SKILL.md
Loading
Related Openclaw Skills
Token Optimizer
A specialized CLI tool for Openclaw Skills designed to analyze, compress, and optimize token consumption in AI agent sessions.
Awwwards-Level Web Design
A specialized skill for building award-winning, immersive web experiences with advanced animations, creative interactions, and high-end visual storytelling.
Context Budgeting
A strategic framework for optimizing context window usage through partitioning and automated checkpointing for long-running agent sessions.
Pollclaw (MeetLark)
A specialized scheduling skill that has evolved into MeetLark to facilitate automated meeting coordination and polling for humans and agents.
Token Counter
A dedicated monitoring tool to track and analyze token consumption across all Openclaw Skills sessions and sub-agents.
ClawFace
A floating, animated desktop avatar that visualizes your AI agent's emotions, tasks, and real-time status.
Featured*
