browse

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

• Third-party content exposure detected (high risk: 1.00). The skill's SKILL.md explicitly supports navigating arbitrary http(s) URLs (goto ) and reading/scraping page content via commands like text, html, links, snapshot and scrape, so it ingests untrusted public web content (third-party pages/social/user-generated content) which the agent is expected to read and that can materially affect subsequent tool actions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill's setup sequence downloads and executes remote installer code at runtime (curl -fsSL "https://bun.sh/install" → saved to a temp file, checksum-checked, then run with bash), so https://bun.sh/install is a runtime-fetched URL whose content is executed as a required dependency.