email

This skill is functionally coherent for an agent that must send and receive email via IMAP/SMTP using curl, and the use of environment-configured inbox credentials matches its stated purpose. However, it grants strong capabilities to an automated agent (read/search/delete mail, send arbitrary messages) and explicitly encourages autonomous action ('Act, don't ask'), creating a high risk of misuse and data exfiltration. Credential usage is direct (username:password passed to curl), and endpoints are arbitrary, allowing connections to attacker-controlled servers if inputs are tampered with. The skill is not obfuscated and contains no direct signs of embedded malware, but its design and operational guidance make it a high-risk capability for an AI agent. Recommend requiring explicit human confirmation for sending/deleting actions, using safer auth methods (tokens, restricted service accounts), auditing and rate-limiting, and avoiding passing credentials on command lines.