bento
Audited by Socket on Nov 9, 2026
4 alerts found:
Securityx3AnomalyThe Bento service and credential-routing patterns are generally consistent with the stated integration purpose, and most scanner findings are documentation artifacts. However, the sandbox instructions explicitly direct unrestricted transfer of credentials and secret-bearing files to remote machines without user approval or redaction. This is a serious, disproportionate exposure risk; classify the skill as SUSPICIOUS, with high security risk but insufficient evidence to conclude confirmed malware.
The shell workflow is a bounded sandbox test and cleanup procedure, with no demonstrated malware or direct credential collection in its commands. However, the accompanying instruction to copy .env files, keys, tokens, and .git into the sandbox could disclose credentials and source code to the provider and should not be followed without explicit authorization and review. No fixed domain, URL, IP address, or credential value appears in the supplied code. Overall severity is medium due to the explicit sensitive-data transfer instruction and the creation of funded, externally managed infrastructure.
The skill is coherent as a remote sandbox operator, but its blanket instruction to copy secrets and credentials without redaction or user confirmation is disproportionate and exposes sensitive data to a provider whose identity and handling are not verified here. No definite credential theft or malicious exfiltration endpoint is shown; classify as suspicious/high risk rather than confirmed malware.
The memory and training actions fit the stated purpose, with no evident concealment or malicious data flow. However, the required Bento CLI’s provenance is unverified in the supplied evidence, so supply-chain risk is elevated under the unverifiable-dependency rule.