skills.rest
silverstein
silversteinCOMMUNITY
Explore all skills by this author →
Website𝕏@matsilversteinGitHub

minutes-video-review

Analyze walkthrough and bug report videos into transcripts, key frames, and actionable briefs.

1.5k|158|Updated Mar 18, 2026
Search Tags:#transcription#bug report#FFmpeg#video analysis#loom#screenpal#key frames
One-click install
npx skills add https://github.com/silverstein/minutes --skill minutes-video-review
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: minutes-video-review
Source: https://github.com/silverstein/minutes/tree/main/tooling/skills/sources/minutes-video-review
Command: npx skills add https://github.com/silverstein/minutes --skill minutes-video-review

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires ffmpeg, yt-dlp, and includes scripts (resource) and references (resource) components.

What problem does it solve?

Recorded demos, bug report videos, Looms, and ScreenPals contain valuable product context, but reviewing them manually is slow and the insights are easily lost. This Skill turns a recorded video into a durable artifact bundle with a transcript, sampled key frames, and a structured analysis that engineering, product, and support teams can act on.

Core Features & Use Cases

  • Minutes-first transcription: Transcribes video audio using the user's existing Minutes setup (Whisper or Parakeet), falling back to hosted captions, local Whisper CLI, or OpenAI transcription only when needed.
  • Adaptive frame sampling: Extracts key frames with ffmpeg using adaptive caps and builds a contact sheet, so long videos do not overwhelm context.
  • Durable artifact bundles: Writes analysis.md, analysis.json, transcript.md, metadata.json, and frames under ~/.minutes/video-reviews/ for follow-up agent work.
  • Use Case: A teammate shares a ScreenPal link showing a bug where pending introductions never appear. Run the Skill on the URL to get a transcript, evidence frames, classified problem signals, and recommended next actions for the engineering team.

Quick Start

Ask the agent to review a video by saying: use Minutes Video Review to analyze this Loom link and return a transcript plus an actionable brief.

Frequently Asked Questions about minutes-video-review

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

FAQPage Schema
How do I analyze a Loom or ScreenPal video with an AI agent?▼

Pass the hosted video URL to the video_review.py script, which downloads it with yt-dlp, extracts audio with ffmpeg, transcribes it, samples key frames, and writes an artifact bundle. The agent then reads the artifacts and produces a brief with issues and next steps.

How do I transcribe a bug report video locally?▼

The skill transcribes locally by preferring hosted VTT captions first, then the Minutes CLI using your existing Whisper or Parakeet configuration, then a local whisper CLI. OpenAI audio transcription is used only as a last resort when configured.

What tools are required to review hosted video URLs?▼

Hosted URLs such as Loom and ScreenPal require yt-dlp for downloading, plus ffmpeg for audio extraction and frame sampling. Local video files work with ffmpeg alone, without yt-dlp installed.

Can I use this skill without the Minutes CLI installed?▼

Yes, the pipeline degrades gracefully. Without the minutes binary it falls back to hosted captions, a local whisper CLI, or OpenAI transcription if OPENAI_API_KEY and the openai CLI are available.

Where are the video review output files stored?▼

Artifacts are written to ~/.minutes/video-reviews/<timestamp>-<slug>/ and include analysis.md, analysis.json, transcript.md, metadata.json, and a frames directory. They are kept separate from the normal ~/meetings/ corpus.

Why does the skill limit the number of sampled frames?▼

Frame caps are adaptive based on video duration to keep context size manageable for the reasoning layer. The transcript serves as the backbone of the review while a curated subset of frames and a contact sheet provide supporting visual evidence.

Related Skills

Explore More Software Engineering Skills→
autonomy-cloudautonomy-cloud
2

screen-recording

Start, stop, pause, resume, and restart screen recording sessions.

Official
Intermediate
VrooliVrooli
54

Swarm Manager Review Agent

Compile plan, diff, and GCT context into verification evidence for backlog items.

Official
Intermediate
beam-ai-teambeam-ai-team

create-meeting-minutes

Transform meeting transcripts into structured Markdown minutes with action items.

Official
Intermediate
ajshedivyajshedivy
4

agno-multimodal

Develop multimodal AI agents that process images, audio, and video.

Community
Intermediate