cover-image

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

• Third-party content exposure detected (high risk: 0.70). The required workflow reads “the blog post or article” (outsider-authored free text) to extract title/key concepts (SKILL.md:25-33), and that extracted prose is used to generate the HTML that is then rendered by headless Chrome, meaning untrusted outsider text can influence the LLM-produced HTML context.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.70). Flagged because the skill advises installing Chrome/Chromium using sudo (e.g., sudo apt install/sudo dnf install) which encourages privilege escalation and also uses Chrome flags like --no-sandbox that weaken sandboxing protections.