designing-email-templates
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to use
curlto upload local files to presigned URLs provided by the vendor API. This is the primary method for image management in the skill's intended use case. - [INDIRECT_PROMPT_INJECTION]: The skill processes external template content, creating an attack surface for indirect prompt injection. 1. Ingestion points: Fetches template data via
workflows-get-email-templateandworkflows-getinSKILL.md. 2. Boundary markers: No explicit separators or 'ignore instructions' warnings are provided for interpolated data. 3. Capability inventory: The agent can execute shell commands (curl) and perform template updates via tool calls. 4. Sanitization: The skill notes that Liquid content is HTML-escaped by default to prevent script injection in emails.
Audit Metadata