The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The example code contains a cross-site scripting (XSS) surface by reflecting potentially user-controlled error messages into raw HTML.
Ingestion points: User data is processed in SKILL.md via req.parse_form::<ContactForm>() within the handle_form handler.
Boundary markers: The example does not utilize delimiters or specific instructions to isolate user-provided error content from the HTML template.
Capability inventory: The skill demonstrates rendering content through res.render(Text::Html(...)) in the show_form handler.
Sanitization: The message value is interpolated directly into an HTML string without escaping, creating a vulnerability surface if the error message contains malicious input.

salvo-flash