email-templates
Pass
Audited by Gen Agent Trust Hub on Jun 13, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [OBFUSCATION]: The skill uses hex-encoded characters (\x3C) to prefix HTML comments in the code snippets within SKILL.md. This technique is typically employed to prevent markdown parsers from rendering the comments as actual HTML during documentation display.\n- [PROMPT_INJECTION]: The template implementation in SKILL.md creates a surface for indirect prompt injection.\n
- Ingestion points: User-provided 'name' and 'confirmUrl' variables are interpolated directly into the WelcomeEmail component in SKILL.md.\n
- Boundary markers: The template lacks delimiters or explicit instructions for the model to ignore any text contained within the dynamic variables.\n
- Capability inventory: The skill includes email delivery capabilities using the Resend API, which could be triggered with malicious content if the model follows instructions injected into the 'name' field.\n
- Sanitization: There is no evidence of input validation or HTML escaping for the interpolated data strings.\n
- Remediation: Implement strict HTML escaping for all user-provided data and wrap untrusted variables in clear boundary markers with instructions to treat the content as data only.
Audit Metadata