A sophisticated security and permission layer providing tiered access control, owner approval workflows, and automated stranger deflection for AI agents.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-access-control
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 agent-access-control 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 Agent Access Control?
Agent Access Control is a critical security framework designed to protect AI agents from unauthorized interactions across various messaging platforms such as WhatsApp, Telegram, Discord, and Signal. By implementing this within your Openclaw Skills setup, you can establish a multi-tier permission system that ensures your agent only interacts with approved individuals while maintaining a professional and diplomatic barrier against unknown senders.
The system focuses on privacy-first communication, allowing the agent to function as a personal assistant without exposing the owner's identity or sensitive tools to the public. Through a structured approval flow, owners can promote contacts through various levels of access, from basic conversation to full administrative control, all while keeping a detailed audit log of interactions.
Agent Access Control Use Cases
- Deflecting unknown senders with a polite, pre-configured automated response
- Managing family and friend access levels for specific agent capabilities
- Protecting sensitive tools and private files from unauthorized execution
- Preventing resource abuse through tiered rate limiting on messaging platforms
- Streamlining the owner approval process for new contact requests via real-time notifications
How Agent Access Control Works
- The skill intercepts every incoming message and extracts the unique sender ID from the messaging platform.
- It normalizes the ID to ensure consistent matching across different formatting styles (e.g., phone number normalization).
- The system checks the ID against the owner list, blocked list, and approved contacts stored in the Openclaw Skills memory.
- If the sender is unknown, it triggers the stranger flow, sending a deflection message and notifying the owner for approval.
- The owner can reply with simple commands like approve or block to instantly update the contact's permission tier.
- On every interaction, the skill enforces tier-specific restrictions, limiting tool usage and data access based on the user's level.
Agent Access Control Setup
To integrate this security layer into your Openclaw Skills environment, create a configuration file at memory/access-control.json with the following structure:
{
"ownerIds": ["your_id_here"],
"approvedContacts": {},
"pendingApprovals": {},
"blockedIds": [],
"strangerMessage": "Hi there! I'm an AI assistant. I've let my owner know you reached out.",
"notifyChannel": "telegram",
"notifyTarget": "your_owner_id"
}
Ensure your owner IDs are correctly formatted for your specific platform (e.g., numeric IDs for Telegram or full international format for phone numbers).
Agent Access Control Data Schema & Taxonomy
The skill manages its state using two primary JSON files within the memory directory to ensure persistence and security:
| File | Purpose | Primary Data |
|---|---|---|
access-control.json |
Permission Configuration | Owner IDs, Approved Contacts, Tiers, Rate Limits |
access-control-log.json |
Audit Trail | Timestamps, Sender IDs, Platform, Actions Taken |
Access is categorized into four distinct levels:
- Tier 0 (Stranger): No access; deflection only.
- Tier 1 (Chat-only): Basic conversation; no tools or memory access.
- Tier 2 (Trusted): Conversation plus public tools like web search or weather.
- Tier 3 (Owner): Full administrative access to all files, tools, and system commands.
Agent Access Control Advanced Features
- Multi-platform ID normalization for seamless identity tracking across Discord, Telegram, and WhatsApp
- Granular rate limiting per access tier to prevent API exhaustion and bot spam
- Real-time owner notification system for immediate approval or rejection of new contacts
- Strict tool-gatekeeping that prevents non-owners from accessing file read/write or shell execution capabilities
- Automatic rotation of audit logs to maintain system performance while keeping a history of access attempts
SKILL.md
Loading
Related Openclaw Skills
OpenClaw Media Gen
A powerful multimedia generation skill that unifies Gemini image and Qwen Wan video creation through a single API key.
OpenClaw LLM Router
A unified gateway providing access to over 70 leading LLMs through a single OpenAI-compatible API.
Model Failover Guard
An automated health monitoring and failover utility that ensures continuous AI agent uptime by switching providers during outages.
Ansible Infrastructure Automation
Automate server provisioning, configuration management, and application deployment using Infrastructure as Code.
Agent Self-Governance
A comprehensive governance framework providing five core protocols to ensure AI agent reliability, verification, and cost-efficiency.
Agent WAL
A Write-Ahead Log protocol that prevents agents from losing critical context and corrections during conversation compaction by persisting state to disk before responding.
METADATA
Featured*
