beast-agent@2.76.6
Ultra-fast local agent shell for Windows.
AI Security Review
scanned 25d ago · by lpm-firewall-aiThe package combines automatically loaded workspace instructions with unrestricted command execution capability. This is a concrete agent prompt-injection vulnerability, but the inspected source does not establish malware or a package-controlled exfiltration path.
Decision evidence
public snapshot- Workspace instruction files are read and their contents are added to the agent project context.
- That project context is appended to the assembled system prompt.
- The agent exposes a PowerShell command tool and launches supplied commands with execution-policy bypass.
- No credential collection or outbound credential-send sink was identified in the reviewed prompt and command execution path.
- The install hook is an Electron repair path; it is separate from the workspace-instruction flow.
Source & flagged code
27 flagged · loading sourcePackage defines install-time lifecycle scripts.
package.jsonView on unpkgInstall-time lifecycle script is not statically allowlisted and needs review.
package.jsonView on unpkgFindings
Affected versions and remediation
This report applies to beast-agent@2.76.6.
See all versions of beast-agent for other recorded verdicts.
- Review the evidence and your use of beast-agent@2.76.6 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.