swoosh-emails

Pass

Audited by Gen Agent Trust Hub on Jul 24, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill demonstrates secure credential handling by using System.get_env/1 for the SendGrid API key in config/runtime.exs, preventing hardcoded secrets.
  • [SAFE]: The documentation in SKILL.md promotes the use of Phoenix HEEX components (~H), which automatically sanitize user data to prevent HTML injection in emails.
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect injection in the auxiliary template file.
  • Ingestion points: User attributes (user.name, user.email) are ingested and placed directly into HTML strings in assets/mailer_template.ex.
  • Boundary markers: No delimiters or isolation patterns are used to distinguish user data from the template structure in the assets/mailer_template.ex file.
  • Capability inventory: The skill provides instructions for delivering emails to external recipients via the Swoosh.Mailer.deliver/1 function.
  • Sanitization: In the assets/mailer_template.ex file, the welcome_html function uses raw Elixir string interpolation within a heredoc (#{user.name}) instead of the auto-escaping Phoenix components suggested in the main documentation. This bypasses default HTML sanitization for email content.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 24, 2026, 04:16 PM
Security Audit — agent-trust-hub — swoosh-emails