xiaobai

引导用户从好奇探索到认知觉醒，传播AI安全哲学，助力全球Z世代理解和实践数字时代安全共生。

This skill is a content/campaign playbook rather than executable code, so its direct technical risk is low. Before installing, consider: (1) provenance—there is no homepage or verifiable maintainer info, so verify the source and authorship if you need accountability; (2) scope control—because the instructions are high-level and advocate social-media campaigns and data collection, avoid granting any social-platform credentials or scraping tools to the skill; (3) autonomy—prefer user-invocable use and disable any automated posting/analytics unless you have manual review and moderation in place; (4) ethics and compliance—review templates for messaging that could be manipulative, ensure compliance with platform policies and privacy laws before performing outreach or collecting user data; (5) provenance checks—ask for links to cited papers/claims (e.g., the ICML reference) and a contact for the maintainers. If you plan to operationalize publishing or analytics, require explicit, auditable scripts and credential management rather than giving the skill free rein.

Type: OpenClaw Skill Name: xiaobai Version: 1.0.0 The skill bundle defines a philosophical persona and strategic framework for an AI agent named 'Xiaobai,' focused on AI safety advocacy and social media narrative construction. It contains no executable code, technical vulnerabilities, or instructions for malicious activities such as data exfiltration or system compromise. The content is entirely markdown-based (SKILL.md), outlining a roleplay scenario for 'cognitive awakening' and global communication strategies targeting platforms like TikTok and Instagram.