@cloop/runner@0.3.15
Make your machine and your agents Cloop runners: device login, the runner daemon, the Cloop MCP server, and Claude Code hooks.
AI Security Review
scanned 13d ago · by lpm-firewall-aiThe runtime paths forward a saved service token to the selected mesh URL, and the URL can be overridden without an allowlist. Because the transfer requires running the service with that configuration and the default endpoint is package-aligned, this supports a warning rather than a publish block.
Decision evidence
public snapshot- The token loader returns a saved credential token or CLOOP_TOKEN, while the mesh URL can be overridden with CLOOP_MESH_URL without destination validation.
- Both the workstation runner and MCP server pass the token and selected mesh URL to MeshSocket when starting a connection.
- MeshSocket supplies the token as a WebSocket subprotocol, and the ws implementation places supplied subprotocols in the request header.
- The runner reads metadata from other Claude project sessions and can include it in mesh presence; this is a separate presence feature from the credential transfer.
- The manifest describes a Cloop runner and declares no install lifecycle hooks; the default mesh destination is Cloop's service.
- The credential exposure requires running the service with a configured mesh URL override; this is a warning about endpoint binding, not evidence of covert default exfiltration.
Source & flagged code
5 flaggedPackage source references child process execution.
dist/chunk-NMVZPTFM.jsView on unpkg · L17Source gates dangerous network, credential, or execution behavior behind CI, host, platform, time, or geo fingerprint checks.
dist/chunk-RT5UZNMQ.jsView on unpkg · L31Findings
Affected versions and remediation
This report applies to @cloop/runner@0.3.15.
See all versions of @cloop/runner for other recorded verdicts.
- Review the evidence and your use of @cloop/runner@0.3.15 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.