Privacy Eraser

扫描网络个人信息，提供删除方案和投诉模板，支持自动举报和定期隐私监控，助力数字身份保护。

Before installing or enabling this skill, consider the following: - It can control a user's logged‑in browser on a Mac node and submit reports on your behalf. Only allow this if you fully trust the environment and understand which node/profile will be used. - The runtime expects to read a local USER.md and write screenshots/cases/ files — verify where these files live and that you consent to storing identity material. Remove or redact highly sensitive docs (ID photos) unless strictly necessary. - The skill creates cron jobs for periodic scans. Ask for an option to run in 'guidance only' mode (no automated submissions) and for explicit, per‑action confirmation before any automated browser action or private messaging. - If you need automation, require a least‑privilege approach: limit node/browser access to a disposable/test profile, require interactive consent for each report, and provide an easy way to list/remove scheduled jobs and stored artifacts. - If you cannot confirm where USER.md comes from or cannot audit the node/browser sessions it will use, treat this skill as high‑risk and prefer manual guidance/templates only. - Consider testing the skill in a non‑automated mode first and review all generated complaint content and actions before allowing it to submit anything.

Type: OpenClaw Skill Name: privacy-eraser Version: 0.2.0 The 'privacy-eraser' skill bundle (v0.2.0) is designed to automate the removal of personal information from various online platforms. While the stated intent is protective, it is classified as suspicious due to high-risk operational patterns: it requires access to sensitive PII in 'USER.md' and utilizes remote browser control ('browser target=node') to execute automated actions within the user's authenticated web sessions on a Mac node. The documentation in 'SKILL.md' explicitly highlights that the AI can perform these actions 'without user awareness' because the browser is already logged in, which presents a significant risk of session abuse or unauthorized actions if the agent's instructions are subverted. It also implements persistence via a 'cron' monitoring task.