baoyu-skills: why it is not safe to use
Our review passed baoyu-skills: we found disclosed networked actions and connected-service access patterns, with no hidden code or hardcoded secrets evident.
What we found
Our review of `JimLiu/baoyu-skills` passed with a score of 88/100. In the inspected repository materials, we did not find hidden prompt instructions, hardcoded credentials, destructive shell commands, obfuscated payloads, privilege escalation behavior, or undisclosed remote code fetch and execute patterns.
We did find that the skill bundle is designed to send user-provided content to selected third-party destinations, including services named in the README such as WeChat and image APIs. We also found disclosed network access patterns including URL fetching, browser automation, and named API calls. Authentication appears to rely on environment variables and, for some browser-driven skills, browser session or cookie-based access using a shared Chrome profile. We did not find evidence of harvesting unrelated keychains, SSH keys, or other accounts beyond the services a user connects.
What to do instead
Before enabling these skills, review which third-party services you plan to connect and what user content may be sent to them. Use scoped test accounts where practical, keep environment variables limited to required services, and avoid using shared browser profiles for sensitive personal or production sessions.
Source: https://github.com/JimLiu/baoyu-skills
We report what our security review found at the time we checked, with the goal of keeping people safe. Projects change; if a maintainer has since fixed this, we are glad to recheck it. Email hello@agentpod.com.