Media Inspector

本地音视频文件分析工具。支持扫描媒体文件、提取元数据、语音转文字（Whisper）、生成摘要和关键片段。支持 MP4/MOV/MKV/MP3/WAV/M4A/FLAC 等格式。

Security Scan

The skill's files and runtime instructions match its stated purpose (local media scanning, metadata extraction, optional Whisper transcription and summary) and do not request unrelated credentials or hidden network endpoints.

Purpose & Capability

Name/description (media scanning, metadata, Whisper transcription, summaries) align with the provided scripts and SKILL.md. The scripts call ffprobe for metadata and optionally use Whisper (CLI or Python) for transcription — these are expected for the described functionality.

Instruction Scope

SKILL.md instructs running the provided scripts against local files and describes outputs. The code only reads filesystem paths the user gives (or recursively scans a given directory), inspects files with ffprobe, optionally loads Whisper models, and writes reports to the specified output directory. There are no instructions to read unrelated system configuration, secrets, or to send data to external endpoints.

Install Mechanism

This is instruction-only (no install spec). SKILL.md recommends installing ffmpeg and Whisper via brew/pip. That's reasonable, but users should be aware installing 'whisper' or loading Whisper models may cause large downloads (model weights). No installer downloads arbitrary code from unknown hosts are present in the package itself.

Credentials

The skill does not request environment variables, credentials, or config paths. The only external resource it may access is Whisper model weights (when using the Python API) which can trigger network downloads; this is proportional to the optional transcription feature and does not require secrets.

Persistence & Privilege

The skill is not always-included and does not modify other skills or global agent settings. It writes output files under the user-specified out-dir and per-file subdirectories only.

Assessment

This skill appears to do what it says: scan local media files, extract metadata with ffprobe, optionally transcribe with Whisper, and write reports. Before running: 1) ensure you install ffmpeg and the Whisper package you prefer and understand that Whisper may download large model files (network activity) when loading models; 2) only point the scanner at directories you intend to analyze (don’t pass your entire home or root if you don’t want a full filesystem inventory); 3) inspect the scripts (they are plain Python) if you have policy concerns — they call subprocesses (ffprobe/whisper) but do not exfiltrate data or call external endpoints themselves; 4) consider running in an isolated environment (container/VM) if you’re worried about untrusted code or large model downloads.

本地音视频文件分析工具，可扫描、提取元数据、转录、摘要和导出报告。

支持格式

视频: mp4, mov, mkv, avi, webm, m4v, ts, mpg, mpeg
音频: mp3, wav, m4a, aac, flac, ogg, opus, aiff, wma

快速开始

1. 扫描媒体文件

python scripts/scan_media.py /path/to/media --out-dir ./media_scan

输出文件：

• scan_results.json - JSON格式
• scan_results.csv - CSV格式
• scan_results.md - Markdown格式

2. 深度分析

python scripts/analyze_media.py /path/to/file.mp4 --out-dir ./media_analysis

分析内容：

1. 元数据提取（ffprobe）
2. 语音转文字（Whisper，可选）
3. 内容摘要
4. 关键片段提取（带时间戳）
5. 报告导出（JSON/CSV/Markdown）

依赖安装

# 必需：ffmpeg (包含 ffprobe)
brew install ffmpeg

# 可选：Whisper 语音转文字
pip install whisper
# 或
brew install whisper

使用规则

• ✅ 优先使用转录文本进行分析，不要仅凭文件名猜测
• ⚠️ Whisper 不可用时，明确告知用户并仅返回元数据
• ❌ 不要虚构内容摘要
• 📝 关键片段必须包含时间戳
• 📊 批量分析时，每个文件单独报告

输出示例

扫描报告

# Media Scan Report

## Summary
- scanned path: /path/to/media
- files found: 10
- audio files: 3
- video files: 7

## Candidates
| filename | duration | type | file size |
|---|---|---|---|
| video.mp4 | 01:30:00 | video | 500.2 MB |
| audio.mp3 | 00:05:30 | audio | 5.2 MB |

分析报告

# Media Analysis Report

## File
- path: /path/to/video.mp4
- type: video
- duration: 01:30:00

## Transcript
- available: yes
- whisper used: yes

## Summary
[基于转录文本的摘要内容]

## Key excerpts
1. [00:01:30 - 00:02:15] 关键片段内容...
2. [00:15:00 - 00:16:30] 另一个关键片段...