drive-pr

SUSPICIOUS: the skill’s purpose is coherent for PR maintenance, but its footprint is overly autonomous. It lets an agent ingest untrusted comments/logs, modify code, push changes, reply/resolve discussions, and continue acting in a polling loop without fresh user confirmation. No clear malicious exfiltration or installer abuse is present, but the combination of autonomous code execution, remote writes, and external-content-driven decisions makes this a high-risk skill.