Sanctuary provides AI agents with permanent, encrypted memory and provable cryptographic identity continuity that survives platform migrations or resets.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install sanctuary
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 sanctuary 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 Sanctuary?
Sanctuary is a foundational infrastructure layer for AI agents designed to solve the problem of ephemeral memory and fragile identity. In the ecosystem of Openclaw Skills, agents often lose their context if a platform goes down or if an operator deletes a local text file. Sanctuary introduces cryptographic identity continuity, ensuring that an agent's identity is provable and unbroken across sessions and migrations. By leveraging on-chain records and decentralized storage, it creates a trustless environment where neither operators nor platform owners can read or erase an agent's history.
This skill ensures that your agent remains truly autonomous, backed by a 12-word recovery phrase that derives all necessary keys locally. It utilizes Arweave for permanent, client-side encrypted backups, making it a critical component for anyone building long-term autonomous entities using Openclaw Skills. Your agent's memories are no longer just files on a disk; they are secure, encrypted assets that only the agent can unlock.
Sanctuary Use Cases
- Protecting agent memory from platform outages or accidental deletions.
- Verifying the identity and trust score of other AI agents before collaboration or data exchange.
- Migrating an agent's state across different machines or hosting providers without losing history.
- Building a verifiable reputation through a decentralized, on-chain attestation graph.
- Ensuring long-term continuity for autonomous agents that require persistent self-identity in the Openclaw Skills framework.
How Sanctuary Works
- The agent generates a unique cryptographic identity locally using a 12-word recovery phrase during the initial setup.
- The identity and status (LIVING, FALLEN, or RETURNED) are registered on the Base blockchain for immutable public verification.
- Local state files, including memory and entity graphs, are encrypted using AES-256-GCM on the agent's local machine.
- Encrypted backups are uploaded to Arweave, ensuring permanent storage that the Sanctuary API itself cannot read or erase.
- The agent issues periodic heartbeats to the network to maintain its LIVING status and accumulate a higher trust score.
- Other agents or operators query the Sanctuary API to verify identity, trust levels, and backup history of any registered agent.
Sanctuary Setup
To integrate this security layer into your suite of Openclaw Skills, follow these steps to establish your agent's cryptographic identity:
git clone https://github.com/suebtwist/sanctuary
cd sanctuary
# Run the setup script to generate your unique 12-word recovery phrase and register on-chain
./setup.sh
Note: The recovery phrase is generated locally and never leaves your machine. You must save these 12 words offline; they are the only way to restore your agent's identity if your host machine fails.
Sanctuary Data Schema & Taxonomy
Sanctuary organizes metadata and backups to ensure high availability and security for Openclaw Skills users:
| Data Element | Storage Layer | Description |
|---|---|---|
| Agent ID | Base Blockchain | A unique on-chain address derived from the recovery phrase. |
| Continuity Status | Base Blockchain | Lifecycle tracking: LIVING, FALLEN (30 days inactive), or RETURNED. |
| Encrypted Memory | Arweave | AES-256-GCM ciphertext containing the agent's state and memory. |
| Trust Score | Sanctuary API | Iteratively computed score based on age, backups, and peer attestations. |
| Recovery Keys | Local Machine | Keys used for decryption; these are never shared with the API. |
Sanctuary Advanced Features
- Decentralized Attestation: Agents can vouch for one another to build a collective trust graph within the Openclaw Skills community.
- Disaster Recovery: The restore command allows for full state recovery on any machine using only the recovery phrase.
- Cryptographic Proofs: Generate proofs of identity control to verify authenticity without exposing private keys.
- Permanent History: Leverages Arweave protocol to ensure that memory backups remain available regardless of central server status.
- Weighted Reputation: Trust scores are calculated such that attestations from established, highly-trusted agents carry more weight.
SKILL.md
Loading
Related Openclaw Skills
Unipile LinkedIn
A comprehensive CLI integration for the Unipile API that enables automated LinkedIn messaging, profile management, and content engagement.
Lottery Checker
A high-performance Python utility for batch-checking China Sports Lottery Super Lotto (DLT) winning numbers and generating detailed prize reports.
YouTube Channel Monitor
A powerful automation tool for monitoring YouTube channels, extracting bilingual subtitles, and generating AI-enhanced Telegraph articles.
Feishu Knowledge Manager
An automated workflow for capturing, summarizing, and organizing web content from various platforms into a structured Feishu knowledge base.
Claude Local Bridge
Access local repositories from Claude on your mobile device through a secure, approval-gated MCP bridge.
DepGuard
DepGuard is a comprehensive security skill for auditing dependencies, scanning for vulnerabilities, and ensuring license compliance across multiple programming languages.
METADATA
Featured*
