Universal Agent Skills
Back to Collection
Tool/Research/Adapted Open Source

Agent Video

Transforms video recordings into structured frames and aligned transcripts.

What It Does

Downloads video through yt dlp, extracts auto scaled keyframes via FFmpeg, retrieves native captions or transcribes with Whisper, and packages timestamps for agent consumption.

Origin and Adaptation Notes

Reviewing lengthy video recordings to extract specific technical diagrams or decisions is time intensive. Timestamped frames enable direct inspection.

Attribution: Adapted from Bradley Bonanno's open source agent-video project. Distributed under the MIT License.

When to Activate

  • •Analyzing recorded conference talks, architectural demos, or lectures
  • •Extracting key slides and visual code samples from video streams
  • •Querying video content via synchronized transcript timestamps

When to Avoid

  • •Simple audio only podcast transcriptions

Skill Specification and Runtime Setup

Full Instruction Specification
# Agent Video Specification

## Overview
Downloads video through yt dlp, extracts auto scaled keyframes via FFmpeg, retrieves native captions or transcribes with Whisper, and packages timestamps for agent consumption.

## Rationale
Reviewing lengthy video recordings to extract specific technical diagrams or decisions is time intensive. Timestamped frames enable direct inspection.

## Core Principles
* Always verify environment constraints and dependencies before executing.
* Enforce negative constraints and reject speculative abstractions.
* Validate output against stated acceptance rubrics.

## Execution Workflow
1. Discovery: Map requirements and inspect active context.
2. Decomposition: Break down the task into discrete, verifiable phases.
3. Execution: Implement surgical changes preserving existing code architecture.
4. Verification: Test all criteria before declaring completion.

Related Skills in Research