profeullerbarros/openclaw-youtube-research-skill

★ 0PythonAudience · researcherComplexity · 2/5ActiveLicenseSetup · moderate

Mindmap

mindmap
  root((youtube-research-skill))
    Inputs
      Video URLs
      Channel feeds
      Watchlist YAML
    Outputs
      Markdown notes
      Weekly briefings
      Verification flags
    Use Cases
      Single video summary
      Channel monitoring
      Recurring research digest
    Tech Stack
      Python
      yt-dlp
      OpenClaw
      YAML

mindmap root((youtube-research-skill)) Inputs Video URLs Channel feeds Watchlist YAML Outputs Markdown notes Weekly briefings Verification flags Use Cases Single video summary Channel monitoring Recurring research digest Tech Stack Python yt-dlp OpenClaw YAML

Things people build with this

USE CASE 1

Summarize a single YouTube video from its captions into Markdown notes inside OpenClaw.

USE CASE 2

Watch a set of channels and get a weekly Markdown briefing of new uploads filtered by topic and language.

USE CASE 3

Flag claims in video transcripts that lack external verification so researchers can chase primary sources.

USE CASE 4

Reuse the YAML watchlist pattern as a template for similar research add-ons.

Tech stack

Pythonyt-dlpOpenClawYAML

Getting it running

Difficulty · moderate Time to first run · 30min

Needs Python 3.10+ and yt-dlp installed; a YAML config must be copied from the example and filled in before first run.

MIT license. You can use, modify, and redistribute the skill for almost any purpose as long as you keep the copyright notice.

In plain English

This project is an add-on, called a skill, for a tool named OpenClaw. The point is to help people do research on YouTube without depending on the YouTube website itself as the main source. It pulls in transcripts, RSS feeds from channels, and metadata, and turns that material into Markdown notes and recurring written reports. The skill can analyze a single video using its captions, watch one or more channels for new uploads, filter what shows up by topics, keywords, and language, and produce weekly or monthly research briefings from a saved watchlist. When something in a video is not externally verified, the output is supposed to mark it as such, so the reader knows it came from the video rather than from a checked primary source. There are also clear limits. The skill does not get around private videos, paywalls, members-only content, DRM, or regional blocks, and it does not download full videos by default. It also does not treat YouTube clips as strong academic evidence on their own. Setup needs Python 3.10 or newer and the yt-dlp tool for fetching public captions. Installation is a git clone into the OpenClaw skills folder, after which the agent should detect it. Configuration is done by copying an example YAML file to a local one and filling in channels, topics, languages, output folders, and an optional delivery target. The local config is kept out of version control so personal paths and channel lists stay private. Once installed, users invoke the skill by name inside an OpenClaw prompt, for a single video, a channel review, or a recurring briefing run. The repository is released under the MIT license.

Copy-paste prompts

Prompt 1

Install the openclaw-youtube-research-skill into my OpenClaw skills folder and configure it to follow three AI research channels in English.

Prompt 2

Run the skill on a single YouTube URL and produce a Markdown summary with claims marked as unverified.

Prompt 3

Set up a weekly recurring briefing across my watchlist, filtered to videos mentioning RAG or evals.

Prompt 4

Sketch how to extend the skill to also pull podcast RSS feeds in the same OpenClaw pipeline.

Prompt 5

Explain why the skill refuses to handle members-only or DRM-protected YouTube content.

Open on GitHub → Explain another repo

Generated 2026-05-22 · Model: sonnet-4-6 · Verify against the repo before relying on details.