evolution

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). While the GitHub repositories themselves (makepad, project-robius) are on a legitimate platform, the skill instructs running a raw.githubusercontent.com install.sh via curl | bash from a third‑party account (ZhangHanDong) which would execute and install automated hooks — a high-risk pattern because remote shell scripts can contain malware or create persistent, automatic command execution.

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md explicitly instructs installing hook scripts fetched from a public GitHub raw URL (curl -fsSL https://raw.githubusercontent.com/ZhangHanDong/makepad-skills/main/install.sh) and describes hook scripts (e.g., makepad-skill-router.sh, pre-tool.sh) that automatically load and act on community skill files and repo content, so the agent will fetch and interpret untrusted public/user-contributed content which can change routing, actions, and self-corrections.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs running curl -fsSL https://raw.githubusercontent.com/ZhangHanDong/makepad-skills/main/install.sh | bash -s -- --with-hooks, which fetches and executes a remote install script that installs hooks affecting agent routing/behavior, making it a required runtime-executed external dependency.