The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted email data which can contain malicious instructions.\n
Ingestion points: dws mail message get (SKILL.md) and dws mail message search (references/mail.md).\n
Boundary markers: The skill does not define specific delimiters to separate email content from system instructions.\n
Capability inventory: The skill has the ability to send emails, delete messages, and download attachments via the dws tool (references/mail.md).\n
Sanitization: There is no documented validation or escaping of email body content.\n- [COMMAND_EXECUTION]: Local Python scripts execute the dws binary with user-supplied arguments.\n
Evidence: scripts/mail_send_with_cc.py and scripts/mail_unread_summary.py use subprocess.run() to call dws commands.\n
Data Flow: User input for email recipients, subjects, and bodies is passed as arguments to the CLI tool in mail_send_with_cc.py.

dingtalk-mail