磁盘垃圾清理大师

磁盘垃圾清理助手。自动扫描磁盘找到老文件、大文件、垃圾文件，生成垃圾桶.md让用户审核确认后再清理。触发词：清理垃圾、清理磁盘、清理硬盘、扫描垃圾、垃圾文件、老文件、大文件、磁盘清理、空间不足、清理缓存、垃圾清理大师

This skill appears to implement a reasonable disk-cleaner workflow, but review these before installing or running: - OS expectation: The SKILL.md and scripts are Windows-focused (C:\ paths, Desktop output). The package metadata does not enforce an OS restriction — avoid running it on non-Windows systems because the deletion routine will directly remove files on non-Windows instead of moving them to a recycle/trash area. - Python path: Examples use C:\Python314\python.exe. Verify a suitable Python interpreter is available and update the command if needed. - Review the generated '垃圾桶.md' carefully before confirming deletion: the deletion script will parse any Windows-style paths found in that markdown and remove them. Ensure lines you want to keep are removed from the file (per the workflow). - Dry-run and backups: Use --dry-run first and save a backup copy of the markdown and any important data before confirming. Prefer moving to Recycle Bin where possible rather than direct deletion; avoid using --no-recycle unless you explicitly want irreversible deletes. - Least-privilege: Run the skill under a user account with only the necessary filesystem access; do not run with elevated/system privileges unless you intentionally need to clean system-wide directories. If you want to proceed, test on a small, non-critical folder first to confirm behavior and ensure the script and SKILL.md assumptions (Windows paths/desktop location) match your environment.

Type: OpenClaw Skill Name: unclecheng-garbage-cleanup-master-cn Version: 1.0.0 The skill bundle provides a disk cleanup utility with a structured workflow involving scanning, user review via a generated markdown file, and final deletion. The scripts (scan_disk.py and delete_trash.py) include safety features such as excluding critical system directories (e.g., C:\Windows), limiting scan depth, and using the Windows Recycle Bin API (SHFileOperationW) for deletions by default. No evidence of data exfiltration, unauthorized remote execution, or malicious prompt injection was found; the tool operates transparently and requires explicit user confirmation before any destructive actions.