html-email-jinja
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes campaign data from external JSON or YAML files to render HTML templates, creating a surface where untrusted data could influence agent behavior if the agent processes the resulting output.
- Ingestion points: Campaign configuration files (e.g.,
config.jsonor YAML) containing subject lines, preheaders, and body copy as described inSKILL.md. - Boundary markers: No specific delimiters are defined for variable interpolation in the templates beyond standard Jinja2 syntax.
- Capability inventory: The skill allows writing rendered HTML to the file system (e.g.,
output/folder) and executing a local Python linter script (scripts/lint_email.py). - Sanitization: The skill provides strong security guidance, recommending the use of Jinja2's
autoescapefeature and cautioning against using the|safefilter for any values supplied by users.
Audit Metadata