base-ui

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

• Third-party content exposure detected (high risk: 0.75). The required workflow explicitly instructs the agent to “fetch current official docs when exact prop names… matter” including downloading Markdown from https://base-ui.com/llms.txt (and related .md docs), which is outsider-authored free text loaded at runtime into the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs the agent to fetch official docs at runtime (e.g., https://base-ui.com/llms.txt and component .md pages such as https://base-ui.com/react/components/popover.md), which would inject external Markdown into the agent's context and directly influence prompts/output.