A secure execution service that allows AI agents to manage EVM wallets and secrets without ever accessing raw private keys.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install safe-skills
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 safe-skills 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 SafeSkills?
SafeSkills is a specialized security framework designed to solve the challenge of credential exposure in autonomous systems. By acting as a secure intermediary, it allows agents to perform sensitive operations—such as signing blockchain transactions—while keeping the raw secrets strictly server-side. This architecture ensures that the agent only handles temporary API keys rather than high-value private keys.
As a robust addition to the Openclaw Skills library, SafeSkills provides a dedicated environment for managing EVM-compatible wallets. It facilitates a trust-minimized workflow where users can monitor and set policies for agent activity through a dedicated frontend, while the agent focuses on executing logic and interacting with decentralized networks securely.
SafeSkills Use Cases
- Deploying autonomous agents capable of managing their own project budgets via EVM wallets.
- Automating micro-transactions and token transfers within Openclaw Skills workflows.
- Interacting with smart contracts securely without storing private keys in the agent's environment variables.
- Providing users with a policy-controlled interface to oversee and limit agent spending.
How SafeSkills Works
- The agent initiates a request to create a secret, such as an EVM wallet, via the SafeSkills API.
- The service generates the secret server-side and returns a restricted API key and a unique claim URL.
- The user accesses the claim URL to authorize the wallet and define spending or security policies.
- The agent uses the API key as a Bearer token to request actions like checking balances or sending transfers.
- SafeSkills validates the agent's request against the user-defined policies and executes the transaction on-chain.
SafeSkills Setup
To start using this with your Openclaw Skills configuration, ensure you have the base API URL set. By default, the service points to the production endpoint.
Initialize a new wallet secret using the following command:
curl -X POST "https://safeskill-production.up.railway.app/api/secrets" \
-H "Content-Type: application/json" \
-d '{
"type": "EVM_WALLET",
"memo": "My agent wallet",
"chainId": 11155111
}'
Store the returned apiKey securely as it is required for all subsequent API calls via the Authorization: Bearer <API_KEY> header.
SafeSkills Data Schema & Taxonomy
SafeSkills organizes data around the concept of a 'Secret'. Each secret contains metadata and specific attributes depending on its type.
| Attribute | Description |
|---|---|
apiKey |
The unique credential used by the agent to authenticate requests. |
claimUrl |
The management link shared with the user to control the secret. |
address |
The public smart account address (for EVM_WALLET types). |
chainId |
The specific blockchain network ID (e.g., 11155111 for Sepolia). |
memo |
A human-readable description of the secret's purpose. |
SafeSkills Advanced Features
- Support for arbitrary transaction execution allows agents to interact with any smart contract via custom calldata.
- Multi-token balance tracking supports both native ETH and any list of ERC-20 token contracts.
- Server-side execution model ensures that raw private keys never enter the agent's context or logs.
- User-in-the-loop policy management through a web-based claim interface.
- Native integration capability with Openclaw Skills for automated financial operations.
SKILL.md
Loading
Related Openclaw Skills
Config Safe
A protective workflow skill designed to safely modify OpenClaw configurations through schema validation and mandatory human-in-the-loop confirmation.
Subagent Watchdog
A reliability tool that monitors AI subagent runs by enforcing completion markers within defined deadlines.
Braingit
A deterministic utility for creating automated git snapshot commits specifically for Markdown files within a repository.
Feishu Agent Relay
A framework for building multi-agent collaboration systems on Feishu by managing cross-bot user identity mapping and task relay.
Trade Manager
An automated trade execution engine for Polymarket that manages risk through stop-loss, take-profit, and trailing stop rules.
Vincent Wallet for AI Agents
A secure, policy-driven wallet infrastructure that allows AI agents to interact with EVM chains and prediction markets without exposing private keys.
Featured*
