setup-agent

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

• Third-party content exposure detected (high risk: 0.80). The SKILL.md's "Scheduled Polling" and OpenClaw Integration sections instruct using OpenClaw cron jobs to periodically check OpenAnt for "unread notifications, submitted tasks," which are user-generated third-party contents the agent would fetch and act on, enabling indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill repeatedly invokes the remote package via "npx @openant-ai/cli@latest", which fetches and executes code from the npm registry at runtime, so it relies on and executes external code fetched remotely.