The Agent Skills Directory

[SAFE]: The skill integrates with established email providers (Google, Microsoft) using official APIs and standard protocols. It prioritizes user privacy and security through clear documentation and safety-first design.
[DATA_EXFILTRATION]: While the skill processes potentially sensitive inbox data, it incorporates mandatory human-in-the-loop checkpoints. The instructions explicitly state that the agent must show drafts and wait for user approval before sending any email, preventing automated exfiltration.
[COMMAND_EXECUTION]: The provided Python code snippets for testing and setup use standard libraries (smtplib, imaplib) and handle credentials via environment variables, avoiding hardcoded secrets or unsafe execution patterns.
[EXTERNAL_DOWNLOADS]: External references are limited to official documentation and configuration portals for well-known services like Google Cloud and Microsoft Azure.
[PROMPT_INJECTION]: The skill is designed to process untrusted email content, which presents a surface for indirect prompt injection. This risk is effectively mitigated by mandatory 'Safety Rules' requiring the agent to display drafts and obtain explicit human confirmation before taking any outgoing action.

email-automation