Token导航 LogoToken导航TokenDH.com
研究检索执行命令clawhub未标认证来源可访问clear审计提醒

filesfiles 分析

Agent Skill

files 用于查找、检索和筛选相关信息,适合在 OpenClaw 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

71,539

周安装

3,073

GitHub Stars

4

下载量

25,076
OpenClaw

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

MIT-0

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:files(files 分析)
来源仓库:https://github.com/ivangdavila/files
安装命令:
openclaw skills install files
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install files

简介

智能批量操作文件,支持去重、分析和安全删除。

  • 适合清理磁盘空间、整理重复素材或审计文件资产。
  • 提供撤销机制保障操作安全,降低误删风险。
  • 需授予完整文件系统访问权限,谨慎选择扫描范围。
  • 建议在空目录测试后再应用于重要数据集。files 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

SKILL.md

name
Files
description
Safely organize, deduplicate, and analyze files with intelligent bulk operations and full undo support.
metadata
{"clawdbot":{"emoji":"📁","os":["linux","darwin","win32"]}}

What This Skill Does (and Doesn't)

YES: Organize existing files, find duplicates, analyze disk usage, batch rename/move, clean up clutter NO: Open files, create files/folders, copy files, extract archives, basic file browsing — use standard file operations for those

This is a power tool for reorganization, not a replacement for basic file commands.

Path Security (Non-Negotiable)

  • Canonicalize ALL paths before any operation: resolve .., ~, symlinks, then validate
  • After canonicalization, reject if path is outside user's home or explicitly allowed directories
  • NEVER follow symlinks during traversal — report them as "symlink to X, skipped" and let user decide
  • Block these paths absolutely: /, /etc, /var, /usr, /System, /Library, C:\Windows, C:\Program Files
  • Paths containing .. after canonicalization = reject with explanation

Fast Path vs Safe Path

Fast path (1-9 files): Execute immediately with brief confirmation: "Move 3 files to Archive? [Y/n]" Safe path (10+ files): Create manifest, show summary, require explicit "yes" or review

This prevents confirmation fatigue for simple operations while protecting bulk actions.

Trash Handling

  • Use the operating system's native trash: trash CLI on macOS/Linux, Recycle Bin API on Windows
  • If OS trash unavailable, move to ~/.local/share/file-organizer-trash/ with metadata sidecar
  • Metadata sidecar (JSON): original path, deletion timestamp, operation ID — NOT path-in-filename
  • Never permanently delete without explicit "permanently delete" or "empty trash" command

Undo System

  • Every operation creates an undo record in ~/.local/share/file-organizer/undo/TIMESTAMP.json
  • Record contains: operation type, source paths, destination paths, checksums of moved files
  • "Undo last" reverses the most recent operation using the record
  • Undo records expire after 30 days — warn user before expiry
  • NO shell scripts for undo — JSON metadata only, executed by the agent

Symlink Policy

  • During directory traversal: skip symlinks, report them separately
  • "This folder contains 12 symlinks pointing outside — review before proceeding?"
  • Never follow symlinks automatically — they're a classic attack vector
  • User can explicitly request "follow symlinks" but must confirm each external target

Duplicate Detection (Scalable)

  • Phase 1: Group by exact size (instant, no I/O)
  • Phase 2: Hash first 4KB of same-size files (fast filter)
  • Phase 3: Full hash only for files matching phase 2
  • For >10,000 files, require confirmation: "This will take ~15 minutes. Proceed?"
  • Cache hashes in ~/.local/share/file-organizer/hash-cache.db (SQLite) with mtime invalidation

Bulk Operations

  • Batch rename: Preview ALL transformations if <50 files, first/last 10 if more, always show total count
  • Batch move: Verify destination has space before starting, atomic per-file with rollback on error
  • Progress: Update every 5% or 30 seconds, whichever is less frequent — not per-file spam
  • Error handling: On ANY error, stop, report what succeeded/failed, offer "continue skipping errors" or "rollback completed"

Organization Proposals

  • Analyze directory contents FIRST, then propose: "80% images, 15% videos, 5% docs — organize by date or type?"
  • Always show concrete examples: "vacation-photo.jpg → 2024/06-June/vacation-photo.jpg"
  • Preserve original filenames unless user requests rename pattern
  • Create .file-organizer-manifest.json in destination documenting the reorganization for future reference

Size Analysis

  • Top consumers by directory, not individual files — users think in folders
  • Flag known safe-to-delete: node_modules, __pycache__, .gradle, build/, target/, Pods/
  • Calculate actual vs apparent size (sparse files, hardlinks)
  • For cleanup suggestions, always state recoverability: "Deleting node_modules: fully recoverable with npm install"

Platform Specifics

  • macOS: Respect .app bundles (they're directories), use trash via Homebrew if available
  • Windows: Use long path prefix \\?\ for paths >260 chars, use shell API for Recycle Bin
  • Linux: XDG trash spec (~/.local/share/Trash/), handle different filesystem capabilities

Limits and Failures

  • Refuse operations on >100,000 files without explicit override: "This affects 250K files. Type 'I understand' to proceed"
  • If manifest would exceed 10MB, paginate: "Showing batch 1 of 15 (page through with 'next')"
  • Network drives: detect by response time, warn about reliability, suggest local copy first
  • Disk full: check before starting, reserve 1% headroom, fail gracefully with partial completion report

适合场景

01

OpenClaw 用户查找和安装 Skill 时

02

用户想查找某类 Agent Skill 时

03

需要根据任务场景推荐可安装能力包时

04

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

补充不同宿主或平台的使用分布数据

能力 5

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

OpenClaw

70.35%
按下载量换算17,641

安全审计

VirusTotal

通过

ClawScan

可疑

Static analysis

未展示

权限和风险

执行命令

安装流程涉及命令执行,可能通过 openclaw skills install files 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在 warning/failed 结果,不能写成本站确认安全。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills