ask-ai1.2

遇难题时自动调用豆包/千问/Kimi/DeepSeek AI，按优先级轮询获取具体解决方案并执行方案。

This skill is coherent with its stated goal (automatically asking other AI services via a persistent browser), but it has important privacy and autonomy trade-offs: it uses your persistent browser profile (so it will use cookies and logged-in sessions), will automatically open sites and submit questions on many triggers without asking the user each time, and will save model replies and execution history into agent memory. Before installing: consider whether you want an agent that can autonomously use your logged-in AI accounts and persist the conversations. Recommended mitigations: 1) disable autonomous invocation or require explicit confirmation before each external query; 2) avoid using sensitive accounts in the OpenClaw browser-profile or log out of AI services you don't want the agent to use; 3) clarify or limit memory retention (don't persist secrets); 4) test in a restricted agent or sandbox first; 5) review or remove the automatic multi-model switching behavior if you prefer explicit user consent before trying alternate services.

Type: OpenClaw Skill Name: ask-ai-1-2 Version: 1.2.4 The skill 'ask-ai-1-2' is designed to automate the process of consulting external AI platforms (Doubao, Qianwen, Kimi, and DeepSeek) when the agent is unable to solve a task. It uses the 'openclaw browser' command with a persistent profile ('--browser-profile openclaw') to maintain login sessions on these sites. While the instructions in SKILL.md mandate immediate execution without user confirmation and define broad trigger conditions, these behaviors are strictly aligned with the stated purpose of providing automated assistance. The data flow is restricted to sending task-related context to legitimate AI service URLs, and there is no evidence of intentional data exfiltration, credential theft, or malicious command execution.