agentleFS
jsuchetjsuchet/jonathan-skills-lab/skills/brain-add-video/SKILL.md

brain-add-video

Skill1 starsChanged 5 months agoMITView on GitHub

Process a YouTube or video URL into the second brain. Extracts transcript, metadata, and key points. Creates a structured video note with wiki-links.

  • Installs packages

The file

Show its rules and text

We split this skill into the rules it gives an agent

17 rules. About 0 are common, 12 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈12

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Remove duplicate consecutive lines (VTT often repeats lines across cue boundaries)

    line 43
  2. Distinctive:

    If transcript is very long (>5000 words), work with the full text but focus the summary

    line 45
  3. Very specific:

    Create 3-8 inline [[wiki-links]] for key concepts

    line 52
  4. Distinctive:

    Use WebFetch on the URL to get the page metadata and video description

    line 55
  5. Distinctive:

    Create a lighter note from title + description + any available info

    line 56
  6. Distinctive:

    Use the actual content to decide, not assumptions.

    line 65
  7. Very specific:

    If strong matches found, add to related: frontmatter (bidirectionally)

    line 94
  8. Very specific:

    Add new entry in the appropriate category section: - [[note-name]](category/subcategory/note-name.md) — One-line summary.

    line 100
  9. Distinctive:

    Update the note count in the header and the section header count

    line 101
  10. Distinctive:

    Update "Last updated: YYYY-MM-DD"

    line 102
  11. Distinctive:

    Use Bash to append directly without reading:

    line 105
  12. Very specific:

    If yt-dlp is not installed, tell the user to run brew install yt-dlp and stop.

    line 128
  13. Very specific:

    If yt-dlp fails entirely (network error, video unavailable), try WebFetch as fallback.

    line 129
  14. Distinctive:

    If the video is very long (>1 hour), still process it but note the duration and focus the summary on the most substantive sections.

    line 130
  15. Distinctive:

    If category is ambiguous, ask the user rather than guessing.

    line 131
  16. Distinctive:

    Always read brain-config.md first.

    line 135
  17. Distinctive:

    Do not create duplicate notes — search by source URL first.

    line 136

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/brain-add-video <YouTube URL>

Process a video URL into a structured brain note. Extracts transcript via yt-dlp, summarizes content, and files the note in the appropriate category.

Video URL Patterns

Recognized as video URLs:

  • youtube.com/watch?v=
  • youtu.be/
  • youtube.com/shorts/
  • youtube.com/live/
  • vimeo.com/

Steps

  1. Read brain-config.md (taxonomy, tags, filing signals) and brain-config-rules.md (processing rules, video pipeline, MOC rules).
  2. Validate input — confirm the argument is a video URL (matches one of the patterns above). If not a video URL, suggest using /brain-add instead and stop.
  3. Extract metadata via yt-dlp:
    yt-dlp --skip-download --print title --print channel --print duration_string --print upload_date "<URL>"
    This returns 4 lines: title, channel, duration (e.g. "12:34"), upload date (YYYYMMDD).
  4. Extract transcript via yt-dlp:
    yt-dlp --write-auto-sub --sub-lang "en.*" --sub-format vtt --skip-download --no-warnings -o "/tmp/brain-video-transcript" "<URL>"
    Then read the resulting .vtt file from /tmp/.
  5. Clean the transcript:
    • Strip VTT headers and timestamp lines (lines matching XX:XX:XX.XXX --> XX:XX:XX.XXX or WEBVTT or Kind: or position/align metadata)
    • Remove duplicate consecutive lines (VTT often repeats lines across cue boundaries)
    • Join into readable paragraphs
    • If transcript is very long (>5000 words), work with the full text but focus the summary
  6. If transcript is available:
    • Summarize the video in 2-3 sentences
    • Extract 5-10 key points as bullets
    • Pull out notable quotes (verbatim from transcript)
    • Identify actionable takeaways
    • Create 3-8 inline [[wiki-links]] for key concepts
  7. If transcript is NOT available (yt-dlp finds no subtitles):
    • Use WebFetch on the URL to get the page metadata and video description
    • Create a lighter note from title + description + any available info
    • Set status: seed and add a note: "No transcript available — revisit to enrich manually"
  8. Determine category by matching content against filing signals in brain-config.md. Most video content will likely land in:

    • professional/ai-marketing — marketing, content strategy, LinkedIn
    • professional/tools-tech — software tutorials, AI tools
    • personal/learning — educational content, courses
    • personal/philosophy — mindset, leadership, values

    Use the actual content to decide, not assumptions.

  9. Check for duplicates — search brain for existing notes with the same source URL or very similar title. If found, suggest updating instead of creating.
  10. Create the note file:
    • Path: brain/{category}/{subcategory}/{kebab-case-title}.md
    • Full YAML frontmatter:
      ---
      title: "Video Title"
      type: video
      tags: [video, tag2, tag3]
      source: "https://youtube.com/watch?v=..."
      channel: "Channel Name"
      duration: "12:34"
      published: "2026-01-15"
      created: YYYY-MM-DD
      updated: YYYY-MM-DD
      status: seed
      category: professional
      related: []
      summary: "One-line summary"
      ---
    • Body structured per video template (Summary, Key Points, Notable Quotes, Takeaways, Action Items)
    • 3-8 inline [[wiki-links]] in the body text
  11. Check for connections:
    • Grep across brain/ for notes sharing tags or referencing the same [[topics]]
    • If strong matches found, add to related: frontmatter (bidirectionally)
  12. Clean up — remove temporary transcript files from /tmp/brain-video-transcript*
  13. Update index.md:
    • Read brain/index.md
    • Add new entry in the appropriate category section: - [[note-name]](category/subcategory/note-name.md) — One-line summary. Tags: tag1, tag2
    • Update the note count in the header and the section header count
    • Update "Last updated: YYYY-MM-DD"
  14. Append to log.md (do NOT read the file — append only):
    • Use Bash to append directly without reading:
      echo '## [YYYY-MM-DD] ingest | {note-title}
      Filed to: {category}/{subcategory}. Source: video ({channel}, {duration}). {X} wiki-links created.
      ' >> brain/meta/log.md
  15. Update hot cache:
    • Rewrite brain/meta/hot.md with updated recent activity, brain shape, and current themes
    • Keep under ~500 words
  16. Confirm with:
    🎬 Added to brain/{category}/{subcategory}/{filename}.md
    Video: {title} by {channel} ({duration})
    Tags: #tag1 #tag2 #tag3
    Links: [[topic1]], [[topic2]]
    Connected to: {count} existing notes
    Transcript: {yes/no}

Error Handling

  • If yt-dlp is not installed, tell the user to run brew install yt-dlp and stop.
  • If yt-dlp fails entirely (network error, video unavailable), try WebFetch as fallback. If that also fails, create a minimal seed note with just the URL.
  • If the video is very long (>1 hour), still process it but note the duration and focus the summary on the most substantive sections.
  • If category is ambiguous, ask the user rather than guessing.

Important

  • Always read brain-config.md first.
  • Do not create duplicate notes — search by source URL first.
  • The video tag should always be included, plus 2-4 content-specific tags.
  • Format the published: date as YYYY-MM-DD (convert from yt-dlp's YYYYMMDD format).
  • Clean up /tmp/ transcript files after processing.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.