laohan-xiazai

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

• Third-party content exposure detected (medium risk: 0.30). The skill workflow involves reading social media feeds, comments, and external videos/posts from platforms like Reddit, YouTube, TikTok, Douyin, Bilibili, and others via tools like agent-reach, opencli, and web search, where an outsider can publish free text that the agent ingests at runtime.

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

• Potentially malicious external URL detected (medium risk: 0.30). The script fetches a binary and certificate from GitHub repository ltaoo/wx_channels_download at setup/runtime via curl, constituting an operational third-party community dependency with common/reputable unofficial provenance.