browse

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

• Third-party content exposure detected (high risk: 0.95). SKILL.md defines a runtime browser workflow (e.g., $B goto <url> then $B text / $B html / $B snapshot) that ingests outsider-authored web page free text/HTML into the agent context via the browser-extracted readable outputs.

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

• Potentially malicious external URL detected (high risk: 1.00). The setup section includes a runtime install command "curl -fsSL https://bun.sh/install | bash" (line 211) which fetches and executes remote code during setup, creating a high-risk runtime dependency.