openhanako-personal-ai-agent

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). While most links are to common services (GitHub, official APIs, localhost), the skill explicitly directs users to download and run unsigned installers from a small/unknown GitHub repo and even instructs bypassing Gatekeeper/SmartScreen—a strong red flag for potential malware distribution.

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md and built-in tools explicitly show agents using a browser/web_search and a "web-scraper" skill that navigates to arbitrary URLs and executes JavaScript (and skills can be installed from GitHub), so the agent is expected to fetch and interpret untrusted public web content which can influence subsequent 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 exposes a runtime installer (skillManager.installFromGitHub) that fetches and installs arbitrary GitHub repositories (e.g. https://github.com/some-user/hanako-skill-weather → https://github.com/) which would download and run remote JavaScript skill code, thus executing remote code and directly affecting agent behavior.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.90). This skill includes explicit instructions and tooling that enable bypassing macOS/Windows security (Gatekeeper / xattr, SmartScreen), disabling safety reviews, and requesting/granting full sandbox/path elevation — actions that can lead to privileged modifications of the host system.