audio-segmenter

当用户想要**把长音频切成小段**、**音频切片**、**音频分割**、**把音频分成固定时长片段**、**制作语音数据集**、**准备Karaoke素材**、**翻唱音频切片**时自动触发。 支持单个音频文件或整个文件夹（支持递归），自动用 ffmpeg 把音频按指定秒数切成小片段，完美保留原始文件夹结构，并智...

Security Scan

The skill does what it says (audio slicing) but contains surprising environment-modifying behavior (automatic pip installs, forced setuptools change, venv creation and restarts, and automated ffmpeg download) that is disproportionate to a simple slicer and should be reviewed before running on a host system.

Purpose & Capability

Name/description match the code: the scripts implement audio slicing, preserve folder structure, and invoke ffmpeg/pydub as described. The heavy dependency and environment management code is broadly related to ensuring ffmpeg/pydub availability, so functionality is coherent — but some of the actions taken (see below) are more invasive than the simple purpose implies.

Instruction Scope

SKILL.md promises automatic venv creation and ffmpeg detection/download; the code does that but also performs additional actions not clearly documented: importing ensure_package triggers immediate pip operations (including forcing setuptools changes) at import time, and the script may restart itself inside a created venv. The code will modify the Python environment and create files/directories beyond the target audio outputs.

Install Mechanism

There is no formal install spec, but the included code performs runtime installs: calling pip (via subprocess) to install packages, and using ffmpeg-downloader to fetch a portable ffmpeg from external hosts (the README mentions gyan.dev, johnvansickle, evermeet). The ensure_package module forces a setuptools reinstall on import — a high-impact operation. Automatic network downloads and pip installs at import/runtime are higher risk than an instruction-only skill.

Credentials

The skill requests no secrets or env vars, but it modifies the host Python environment: ensure_package.fix_setuptools_for_legacy_packages runs pip to reinstall setuptools at import time; ensure_package.pip installs packages into the current interpreter (which may be the system Python) before the code switches to a created venv. The venv path is created outside the skill folder (VENV_DIR points to a parent-level 'venv'), which is unexpected. These operations are disproportionate for a simple slicing task and could affect unrelated projects.

Persistence & Privilege

always:false and no special platform privileges, but the skill creates persistent artifacts: a venv directory, downloaded ffmpeg binaries, log files under SKILL_ROOT/logs, and potentially installed packages in the current environment. It also restarts the process inside the venv, which increases persistent footprint. This is expected for a script that self-manages dependencies but is worth user consent.

What to consider before installing

This skill appears to implement legitimate audio-slicing, but it will automatically install packages and modify your Python environment (including forcibly reinstalling setuptools), create a virtual environment in a parent path, download ffmpeg, write logs, and restart itself inside the venv. Before installing or running: 1) Review ensure_package.fix_setuptools_for_legacy_packages and ensure_package.pip to confirm you accept automatic pip operations. 2) Run the tool inside an isolated environment (container or throwaway virtualenv) or as a non-privileged user to avoid affecting system Python/site-packages. 3) If you only want slicing, consider running the code manually after creating and activating a venv yourself (set RUNNING_IN_VENV=true to avoid automatic venv creation), or modify the script to remove the forced setuptools/pip behavior. 4) Do not run as root/Administrator. If you are uncomfortable with on-import pip activity or unexpected filesystem writes, treat this skill as suspicious and do not install it on production/workstation systems.

功能：一键把长音频切成固定时长的片段，专为语音训练、素材整理、翻唱/Karaoke 准备、数据集制作设计。完美保持原文件夹结构 + 智能默认输出路径 + 自动处理 ffmpeg。

触发时机（Triggers）

• 用户提到“音频切片”“切成小段”“分割音频”“切片”“分段”“每段XX秒”等词，并提供音频文件路径或文件夹。
• 用户说“帮我处理这个长音频”“制作语音数据集”“Karaoke素材准备”等场景。
• 支持单个文件或整个文件夹（可递归）。

支持的音频格式

.mp3 / .wav / .m4a / .ogg / .flac / .aac / .wma 等（pydub 支持的所有常见格式）

参数说明（默认值）

1. -i → 输入路径（必填，支持单个文件或文件夹）
2. -d → 每段切片时长（秒），默认 60
3. -o → 输出目录（不传则智能选择：单文件同目录，文件夹则在同级创建 [输入文件夹名]_sliced_audio）
4. -r → 文件夹模式下是否递归子文件夹（默认否）

参数提取指南

当决定调用此技能时，请从用户上下文中提取以下参数：

1. <输入路径> (必填): 用户提供的音频文件路径或文件夹路径。
2. <切片秒数> (选填): 用户指定的每段时长（秒），未提则默认 60。
3. <输出目录> (选填): 用户指定的输出路径，未指定则使用智能默认。
4. 递归标志 (-r): 用户提到“递归”“所有子文件夹”时添加。

执行步骤

1. 解析路径：自动识别用户的输入是单个音频文件还是文件夹。

2. 默认输出：未指定 -o 时，单文件输出到源目录附近，文件夹则在同级创建带 _sliced_audio 的新文件夹，并保留原始目录结构。

3. 调用命令：使用以下兼容性命令启动脚本（优先 python3，失败则 python）。脚本会自动创建虚拟环境、检测并安装 ffmpeg。

   (python3 scripts/audio_slicer.py -i "<输入路径>" [-d <切片秒数>] [-o "<输出目录>"] [-r]) || (python scripts/audio_slicer.py -i "<输入路径>" [-d <切片秒数>] [-o "<输出目录>"] [-r])