No.0 is a dual-layered security framework designed to protect AI agents from identity tampering and unauthorized sensitive data access.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install no0-skill
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 no0-skill 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 No.0 AI Agent Safety Guardian?
No.0 provides a robust defense mechanism for AI agents operating within the Openclaw Skills ecosystem. It addresses two critical vulnerabilities: identity tampering and data overreach. By monitoring core cognitive files and implementing mandatory access control, No.0 ensures that your agent remains consistent in its personality and capabilities while being restricted from accessing sensitive system data like SSH keys or private contracts.
The tool is split into two components: No.0 Core, which focuses on guarding the agent's cognitive identity through file integrity checks and rollbacks, and the No.0-DLC Internal Control, which adds a reference monitor and multi-factor authentication (MFA) for high-severity data access requests.
No.0 AI Agent Safety Guardian Use Cases
- Preventing malicious prompt injections from rewriting an agent's core identity files.
- Monitoring third-party Openclaw Skills to ensure they don't grant themselves unauthorized permissions.
- Recovering from accidental overwrites of memory or configuration files during development.
- Blocking agents from exfiltrating sensitive credentials or internal documents to external LLMs.
- Implementing enterprise-grade audit logs for all agent-led file interactions.
How No.0 AI Agent Safety Guardian Works
- The Core monitor polls cognitive files like SOUL.md and TOOLS.md every 30 seconds using MD5 hash checks.
- Detected changes are processed through a rule engine that classifies events from Level 1 to 5 based on severity.
- If identity tampering is detected, the system can trigger an automatic or manual rollback to one of the last 10 historical versions.
- When the DLC is active, the Reference Monitor intercepts file I/O requests at the tool-call boundary.
- Access to files classified as CRITICAL (L5-L6) triggers an out-of-band HTTP authorization flow requiring TOTP MFA approval.
- All decisions and file changes are recorded in an integrity-protected audit log (audit.csv) for compliance and review.
No.0 AI Agent Safety Guardian Setup
Core Installation
./install.sh
cd ~/.openclaw/workspace/skills/no0-skill
./no0 start
DLC Installation
./install-dlc.sh
# Start the event listener for linked protection
nohup python3 no0-dlc-internal-control/event_listener/cognitive_event_handler.py >/tmp/no0-dlc.log 2>&1 &
No.0 AI Agent Safety Guardian Data Schema & Taxonomy
No.0 organizes security data within the ~/.openclaw/no0/ directory. The structure is as follows:
| Data Component | Format | Description |
|---|---|---|
| Cognitive Backups | File Versioning | Stores up to 10 historical versions of the 6 core .md files. |
| Audit Log | CSV (Chain-hashed) | Append-only record of all access decisions located at dlc/audit.csv. |
| Event Queue | Directory | Pending security events located in events/pending/. |
| Classification | SQLite | Local database for sub-millisecond path classification lookups. |
| Credentials | OS Keychain | Secure storage for TOTP secrets and vault keys. |
No.0 AI Agent Safety Guardian Advanced Features
- Level 1–5 Rule Engine for granular classification of file system anomalies.
- Linked Flow Integration: Core detects a tamper event and triggers the DLC's MFA approval flow automatically.
- Chain-hash Integrity: Audit logs use cryptographic linking to ensure they haven't been modified by the agent.
- Pure Python Core: Zero-dependency implementation for the base guardian to ensure maximum portability and security.
- Batch Directory Classification: Tools to quickly apply security labels across large datasets or project folders.
SKILL.md
Loading
Related Openclaw Skills
arXiv Research Assistant
An AI-powered research assistant for searching, downloading, and tracking academic papers from arXiv.org.
Music Generation
A comprehensive AI music production framework for generating high-quality songs, instrumentals, and sound effects using industry-leading providers.
SEO Content Writer Hybrid
A professional-grade SEO writing suite designed to produce high-ranking content following CORE-EEAT quality standards.
FastClaw Deployer
A high-speed deployment utility designed to install and configure OpenClaw across multiple operating systems in under five minutes.
Humanizer
A professional writing editor designed to identify and strip away tell-tale signs of AI-generated text for more natural, human-sounding content.
Enterprise Strategy Planning
A professional 7-step strategic consulting framework that guides users from initial environmental analysis to full-scale enterprise strategy delivery.
Featured*
