parallels-vm

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

• Third-party content exposure detected (high risk: 1.00). The skill's required workflows and scripts (e.g., scripts/prl-macos-install-openclaw.sh, scripts/prl-linux-install-openclaw.sh, and scripts/prl-windows-install-openclaw.sh) explicitly download installers from https://openclaw.ai (and accept arbitrary --install-url/--spec URLs) and the various prl-*-openclaw-update-verify.sh and gateway-status scripts parse that fetched output/JSON to decide follow-up actions, so untrusted third‑party web content is ingested and can materially influence tool use and decisions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill's install/update wrappers (e.g., prl-macos-install-openclaw.sh, prl-linux-install-openclaw.sh, prl-windows-install-openclaw.sh) download and then execute remote installer scripts at runtime from https://openclaw.ai/install.sh (and https://openclaw.ai/install.ps1), meaning fetched content is executed in guests and is relied on by the skill as a required installer/update dependency.