review-package

The skill’s purpose is coherent and its visible behavior is mostly local packaging, not credential theft or covert exfiltration. However, its core action depends on executing an unverifiable local script and an undocumented analyzer component, so the overall classification is suspicious/high-risk from an execution-trust standpoint rather than confirmed malware.

The code does not show clear malware, exfiltration, credential theft, or backdoor behavior. It is a legitimate packaging script with insufficient path validation. Untrusted file lists can read outside the project and write outside the package, while OUTPUT_NAME can influence deletion and archive locations. On macOS, OUTPUT_NAME also creates a potential AppleScript injection issue. Inputs should be canonicalized and constrained to expected directories, OUTPUT_NAME should be validated as a basename, and the AppleScript path should be safely escaped.