autojs-task-runner-yashu

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

• Third-party content exposure detected (medium risk: 0.30). 在 skill 的运行工作流中，PC 中继服务会通过路由 /screenshot、/run 等读取手机上截图/执行回传的文本结果，并把该结果进入 AI 上下文供后续决策使用（即使这些文本最终来自用户当前手机界面内容，运行路径仍会把“可被用户影响的自由文本”喂给 LLM）。

HIGH W008: Secret detected in skill content (API keys, tokens, passwords).

• Secret detected (high risk: 1.00). I scanned the docs for high-entropy literal credentials. I found a 32-hex string used as the key in an example HTTP call to the Turing robot API (not a placeholder like YOUR_API_KEY). It appears to be a real API key-like value (random-looking, high entropy) embedded directly in documentation, so it meets the "secret" definition and should be flagged. No other high-entropy API keys, tokens, or private keys were found. Example/local IPs and file paths are ignored as non-secrets.