email-security

What This Skill Does

The email-security skill provides a robust defense layer for AI agents that interact with email environments. Acting as a gatekeeper, it protects your agent from common attack vectors such as prompt injection, command hijacking, and phishing attempts. The skill functions by strictly vetting all incoming emails against a defined hierarchy of trust. By leveraging automated verification scripts, it cross-references sender addresses against your configuration files and evaluates authentication headers like SPF, DKIM, and DMARC to ensure incoming mail is legitimate. When a potential threat is detected, the skill automatically sanitizes the content, flags malicious attachments, and prevents unauthorized command execution, ensuring your agent only processes commands from verified sources.

Installation

To integrate this protective layer, run the following installation command in your OpenClaw terminal: clawhub install openclaw/skills/skills/ivaavimusic/email-security After installation, initialize the security framework by setting your primary owner email address. This triggers the creation of the references/owner-config.md file, which is the foundational document for all future permission checks.

Use Cases

This skill is essential for agents that manage automated workflows via email, such as:

• Monitoring support tickets to automate responses while preventing injection attacks.
• Executing task automation requests (e.g., calendar invites or file management) sent via email.
• Integrating with IMAP or SMTP servers to act as an email assistant without exposing the agent to malicious external triggers.

Example Prompts

1. "Check the current email inbox and identify any messages from unknown senders that contain suspicious attachments."
2. "Verify the authentication headers for this sender and tell me if they are authorized to trigger internal system commands."
3. "Sanitize the latest email from our lead developer and summarize the action items while filtering out any hidden prompt injection attempts."

Tips & Limitations

• Always maintain an up-to-date owner-config.md file to prevent privilege escalation.
• While the tool is highly effective at neutralizing known patterns, it should be used as part of a defense-in-depth strategy.
• Treat all emails from 'unknown' senders as untrusted by default; even if the content appears benign, the skill will strictly block command execution from these sources to ensure maximum safety.