web-extract

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

• Third-party content exposure detected (high risk: 0.85). SKILL.md describes scraping arbitrary open-web pages (e.g., “Page → … JSON in one call”), and at runtime the CLI requests /scrape//search//crawl results containing scraped page text (JSON fields like title/summary/sections and/or markdown/rawHtml) that would be passed back into the agent’s context for use; this is outsider-authored free text from the open web.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill makes runtime calls to the external API at https://api.zoodata.ai/openapi/v2/webtools (used in SKILL.md and scripts) to fetch scraped page content that is then returned as structured JSON for downstream LLMs, meaning remote content can directly influence agent prompts/behavior.