The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides a migration template for a MailServlet in SKILL.md that ingests untrusted data from HTTP request parameters and interpolates it into email headers and body without sanitization, creating an attack surface for indirect prompt injection or mail header manipulation.
Ingestion points: The doPost method in the MailServlet snippet (file SKILL.md) calls request.getParameter for to, subject, and body fields.
Boundary markers: The provided code snippets do not include boundary markers or explicit instructions to the agent to treat this data as untrusted or to ignore embedded instructions.
Capability inventory: The skill provides instructions to implement a network-capable email transmission flow using Transport.sendMessage and custom SOCKS5 proxy logic in ConnectivitySocks5ProxySocket.java and OnPremiseSMTPTransport.java.
Sanitization: The example code lacks validation, escaping, or filtering of external content before it is processed by the mail session.

mail-destinations