skills.rest
reason-machines
reason-machinesOFFICIAL
Explore all skills by this author →
Website𝕏@reasonmachinesGitHub

watch-cli-video-agent

Extract frames and transcripts from social videos for AI agent analysis.

4|Updated May 16, 2026
Search Tags:#AI agent#youtube#transcription#yt-dlp#ffmpeg#video analysis#frame extraction
One-click install
npx skills add https://github.com/reason-machines/devtools-skills --skill watch-cli-video-agent-reason-machines
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: watch-cli-video-agent
Source: https://github.com/reason-machines/devtools-skills/tree/main/skills/watch-cli-video-agent
Command: npx skills add https://github.com/reason-machines/devtools-skills --skill watch-cli-video-agent-reason-machines

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires yt-dlp, ffmpeg, jq, curl, python3.

What problem does it solve? AI agents cannot natively watch videos, and calling multimodal video APIs on full-length content is expensive and slow. This Skill decomposes videos from YouTube, X/Twitter, LinkedIn, TikTok, Reddit, Vimeo, and Facebook into JPG frames plus a text transcript, letting agents reason about video content with standard vision and text models at roughly 50x lower cost. ## Core Features & Use Cases - Full video pipeline: The watch command downloads a video, extracts evenly-spaced frames, and transcribes audio in one structured output block. - Modular commands: Use dl-video, extract-frames, transcribe, and audio-q individually for download-only, frame extraction, Whisper transcription, or audio scene Q&A. - Login-walled content: Auto-detects browser cookies or accepts a --cookies file for LinkedIn, private X, and Facebook videos. - Use Case: Ask the agent to watch a coding tutorial video; it reads frames showing file structures and code on screen, combines them with the transcript's verbal instructions, and reconstructs a working implementation. ## Quick Start Ask the agent to watch a YouTube tutorial video at a given URL and summarize what it teaches using the extracted frames and transcript.

Frequently Asked Questions about watch-cli-video-agent

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I analyze a YouTube video with an AI agent?▼

Run the watch command with the video URL to download it, extract evenly-spaced frames, and transcribe the audio. The agent then reads the frames as images and the transcript as text to reason about the full video content.

What tools extract frames and transcripts from videos?▼

watch-cli combines yt-dlp for downloading, ffmpeg for frame and audio extraction, and Whisper Large v3 Turbo via the Kyma API for transcription. It wraps these into single commands like watch, extract-frames, and transcribe.

Can I download login-walled LinkedIn or private X videos?▼

Yes, the tool auto-detects cookies from signed-in browsers like Chrome, Firefox, Safari, Edge, and Brave. For servers or CI, export a cookies.txt file with a browser extension and pass it via the --cookies flag.

How many frames should I extract from a video?▼

Use 4-8 frames for short clips under 2 minutes, 8-16 for tutorials, 16-24 for lectures, and 24-32 for conference talks over an hour. Double the recommendation for fast-cut UI demos to catch rapid edits.

Why does transcription fail on long videos?▼

The transcription provider enforces a 25MB audio limit, so videos over two hours may be rejected. Split the video with ffmpeg into smaller segments, transcribe each part separately, and concatenate the resulting transcripts.

How much does video transcription cost compared to multimodal APIs?▼

Transcription costs about $0.04 per hour of audio, and frame extraction is free since it runs locally with ffmpeg. A one-hour video costs roughly $0.10 total versus about $5 with a multimodal video API.

Related Skills

Explore More Software Engineering Skills→
mathiaschumathiaschu
92

watch

Download videos and convert them into timestamped frames and transcripts.

Community
Advanced
nicepkgnicepkg
276

mermaid-tools

Extract Mermaid diagrams from markdown files and generate PNG images.

Official
Intermediate
AradotsoAradotso
4

watch-cli-video-agent

Decompose videos into frames and transcripts for agent analysis.

Official
Intermediate
LeooooooowLeooooooow
7

audio-to-text-caption

Transcribe audio into caption-ready text with automatic speech recognition and formatting.

Community
Intermediate