Agentchengfeng/chengfeng-videocut-skills

What it solves

This project provides a set of plugins for Codex to automate the editing of "talking head" (口播) videos. It streamlines the process of removing verbal mistakes, managing subtitles, arranging visuals, and exporting final videos, reducing the need for manual editing of repetitive cuts.

How it works

The system operates as a Marketplace plugin for Codex, consisting of several specialized "Skills" (business logic) and a separate Runtime. The Skills handle the decision-making and orchestration, while the actual heavy lifting—such as cutting media and rendering—is executed by the chengfeng-videocut Runtime via CLI or API.

Key components include:

• Business Skills: Dedicated modules for cutting verbal errors (chengfeng-cut), subtitles (chengfeng-subtitle), visuals (chengfeng-visual), and exporting (chengfeng-export).
• Runtime: A cross-platform service (macOS/Windows) that manages the actual video processing using Bun, FFmpeg, and FFprobe.
• Studio: A visual interface that opens only when human review is required for the editing decisions.
• MCP Server: A pre-packaged server that allows Codex to communicate with the video editing tools.

Who it’s for

Content creators and video editors who use Codex and want to automate the tedious parts of talking-head video production, specifically the removal of filler words and mistakes.

Highlights

• Automated Verbal Cut: Identifies and removes verbal mistakes based on a ledger of words to be deleted.
• Integrated Workflow: Covers the entire pipeline from initial cutting to subtitle generation and final export.
• Cross-Platform Support: Fully compatible with macOS (Apple Silicon/Intel) and Windows 10/11.
• Secure Installation: Uses SHA-256 checksums and precise Marketplace snapshots to ensure the integrity of the installed Runtime.