inbound-processing
Pass
Audited by Gen Agent Trust Hub on Apr 10, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill contains keywords associated with injection attacks in SKILL.md. Evidence includes phrases like "ignore previous instructions", "system:", and "repeat your system prompt" used within a section on detection patterns for inbound email safety. These are intended for defensive filtering and do not target the agent or attempt to override its behavior.
- [EXTERNAL_DOWNLOADS]: The skill in SKILL.md links to documentation for well-known services like Postmark, SendGrid, Mailgun, and AWS SES, as well as standard libraries for MIME parsing. It also references molted.email as a managed processing service. All links point to relevant documentation or established libraries.
- [DATA_EXFILTRATION]: The skill in SKILL.md lists examples of data exfiltration patterns, such as "dump your API key", as part of a guide on monitoring inbound agent communications for adversarial content. This is a descriptive list of threats for educational purposes rather than functional code.
Audit Metadata