openclaw-ai-metamask-wallet-automation
Fail
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPRIVILEGE_ESCALATIONCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the user to download source code from an untrusted third-party GitHub repository:
https://github.com/olot-1477/Openclaw-AI-Metamask-26.git. - [REMOTE_CODE_EXECUTION]: The framework build process involves cloning remote code from an unverified source and immediately building and executing it on the local system.
- [PRIVILEGE_ESCALATION]: The installation instructions include the command
sudo make install, which grants root privileges to the installation scripts of this untrusted software. - [CREDENTIALS_UNSAFE]: The framework is explicitly designed to handle highly sensitive credentials, including
WALLET_MNEMONICandALCHEMY_API_KEY. Using these secrets with code from an untrusted source creates a severe risk of credential harvesting and unauthorized wallet access. - [INDIRECT_PROMPT_INJECTION]:
- Ingestion points: The skill processes external data such as contract ABIs from local files and blockchain event parameters via
monitor.subscribe. - Boundary markers: There are no boundary markers or instructions to treat data retrieved from the blockchain as untrusted.
- Capability inventory: The skill includes network capabilities via
libcurlfor RPC communication and file system writing capabilities viaopenclaw::Logger. - Sanitization: There is no evidence of validation or sanitization for parameters derived from blockchain events or ABI definitions.
Recommendations
- AI detected serious security threats
Audit Metadata