The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from external emails, which introduces a surface for indirect prompt injection.
Ingestion points: The skill processes raw email headers, Received chains, and external URLs (SKILL.md).
Boundary markers: The instructions include a 'Mandatory confirmation gate' that requires user confirmation before taking action on targets.
Capability inventory: The skill references external tools such as dig, nslookup, and urlscan for investigation. No internal scripts or automation are present.
Sanitization: The skill lacks instructions for sanitizing or escaping untrusted input before processing.
[NO_CODE]: The provided skill contains only documentation and checklists in markdown format and does not include any source code, scripts, or binary files.

email-security