@adityaaria/spark@6.0.21

No blocking static signals were detected.; previous stored version diff introduced dangerous source

User runs `spark dashboard` or `spark ui`.

A reachable webpage or network client could write a project skill file, potentially escaping the intended skills directory via path traversal.

CORS-enabled dashboard POST writes arbitrary skill content using an unsanitized path component.

The package is primarily an agent skills/plugin distribution. A user-invoked dashboard command starts a CORS-permissive HTTP server and lets POST /api/skills write supplied content under a path derived from unsanitized data.name; because the server does not pin listen() to loopback, this creates a real write-surface if the dashboard is reachable. I did not find install-time execution, credential collection, or exfiltration that would support a malicious verdict.

Static inspection supports a suspicious warning for a dashboard write vulnerability, not a publish block for malware. The scanner's dangerous-delta hint maps to package-aligned dashboard code, but that code still exposes an unresolved arbitrary skill write risk.