agent-update-checker

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

• Third-party content exposure detected (high risk: 0.75). SKILL.md:16-17 describes the runtime workflow where the agent fetches and analyzes content from a public GitHub repository URL (outsider-authored web page/README text) and then uses it to generate an update report in the LLM context.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly fetches and may install code at runtime from the repository URL https://github.com/lensetek/Research-Agent-Skills-Collection, which can result in executing remote code and directly change agent behavior.