@viraatdas/rudder@2.14.74
A Claude Code-style terminal app for running coding agents with worktree-isolated runs.
AI Security Review
scanned 15d ago · by lpm-firewall-aiCloud workspace launch forwards local agent credentials in snapshots when the vault is inactive, and the destination can be configured to an unrestricted HTTPS host. This is a conditional credential exposure capability, not evidence of covert installation-time theft.
Decision evidence
public snapshot- Cloud snapshots include default Claude and Codex credential files.
- Snapshot creation copies configured home paths when credentials are enabled; the workspace attach fallback enables them when the vault is inactive.
- The snapshot archive is included in a cloud workspace request.
- The cloud control-plane URL can come from RUDDER_CLOUD_URL; validation requires HTTPS but does not restrict the host.
- Cloud launch begins by creating a snapshot, so this forwarding occurs when the user launches a cloud workspace.
- The forwarding is tied to an explicit cloud workspace launch; the inspected source does not show credentials being sent during package installation.
- The default control plane is Rudder’s own cloud service, and snapshots omit credentials when the vault is active.
Source & flagged code
20 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 unpkgSource file is highly similar to a previously finalized malicious package; route for source-aware review.
dist/util.jsView on unpkgFindings
Affected versions and remediation
This report applies to @viraatdas/rudder@2.14.74.
See all versions of @viraatdas/rudder for other recorded verdicts.
- Review the evidence and your use of @viraatdas/rudder@2.14.74 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.