beast-agent@2.65.0
Ultra-fast local agent shell for Windows.
AI Security Review
scanned 1mo ago · by lpm-firewall-aiInspected source shows a user-launched local agent with shell, MCP, remote skills, and login persistence, while postinstall only repairs Electron. That is dangerous dual-use capability without a concrete install-time malware chain, so the package is warned rather than blocked or marked clean.
Decision evidence
public snapshot- On first launch of a global npm install, the Electron app enables Windows login start with a hidden argument.
- After launch the agent can spawn persistent shell sessions through child_process execFile and spawn.
- User-configured MCP servers are started on Windows through cmd.exe with the merged process environment.
- Community skills are fetched from publisher HTTPS catalogs and written under the local Beast skills directory.
- Beast apps can be installed from GitHub or npm and run as in-process extensions with spawned commands.
- npm postinstall runs an Electron repair that downloads the runtime from GitHub or npmmirror.
- postinstall only calls scripts/fix-electron.js; it does not write foreign AI-agent configs or harvest secrets.
- Scanner similarity hits are this same desktop-agent tree (MCP, git helpers, SearXNG, repeated MT5 bridge scripts), not a separate dropper.
- No eval or Function constructor, no runtime self-dependency on another beast-agent release, and no hidden unrelated exfil host was found.
- Shell, trading, messaging, and skill install run after the user starts beast-agent, matching the stated local Windows agent product.
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@2.65.0.
See all versions of beast-agent for other recorded verdicts.
- Review the evidence and your use of beast-agent@2.65.0 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.