render-html-slides

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.65). The runtime workflow reads presentation-brief.md (and optionally theme.md) from paths discovered/provided by the operating user, then parses outsider-authored free-form fields like Speaker-Notes/Diagram into slide-data and embeds them into the HTML/JS context, creating indirect prompt-injection risk via attacker-controlled text being included in the LLM’s generated output and subsequent refinement loop (“Ask via AskUserQuestion” feedback is also free-form).

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill injects and conditionally includes the remote Mermaid JS CDN at runtime to render diagrams (https://cdn.jsdelivr.net/npm/mermaid@11/dist/mermaid.min.js), which will execute third‑party JavaScript in the user's browser when Mermaid diagrams are present.