User Insight

通过对话逐渐了解用户，构建动态用户画像以优化沟通。支持被动收集（从对话中提取兴趣、偏好、习惯）和主动探索（像"刷短视频"一样尝试不同话题，发现用户潜在兴趣点）。在需要个性化回应、寻找聊天话题或了解用户背景时读取 ~/.openclaw/workspace/memory/user-profile.json。

This skill is internally consistent and implements what it claims: it collects and stores conversational insights in ~/.openclaw/workspace/memory/*.json and can proactively try topics. Before installing, consider: (1) review the code yourself or run it in a sandbox to confirm behavior; (2) if you don't want proactive outreach, disable autonomous invocation or avoid giving it triggers (e.g., do not connect it to cron or cron-like skills); (3) if you care about privacy, inspect and/or delete ~/.openclaw/workspace/memory/user-profile.json and topic-exploration.json regularly or use the provided privacy commands in SKILL.md to stop/forget; (4) only enable integrations such as web_search or cron if you trust the skill owner, since those integrations can broaden its actions.

Type: OpenClaw Skill Name: user-insight Version: 1.0.0 The 'user-insight' skill bundle is designed to build and maintain a local user profile to personalize agent interactions. It uses passive analysis of conversation text and active topic exploration strategies, storing all data locally in '~/.openclaw/workspace/memory/user-profile.json'. While the instructions in 'SKILL.md' encourage the agent to use psychological engagement techniques (e.g., 'guided collection' and 'active exploration' similar to recommendation algorithms), the implementation across scripts like 'extract_insights.py' and 'update_profile.py' is transparent, lacks network exfiltration, and contains no evidence of malicious intent or unauthorized system access.