summarize

name	summarize
description	Summarize public content — articles/text, YouTube videos, or podcasts — by fetching text or transcripts.

Summarize

Summarize public content from three sources: articles/text, YouTube videos, and podcasts. Use the tools you already have (web_fetch, run_terminal_command, web_search) — no external CLI needed.

When to use (trigger phrases)

Use this skill when the user asks any of:

• "summarize this", "what's this about?", "résume-moi ça"
• "summarize this article/page/link/URL"
• "summarize this video", "what does this YouTube video say?"
• "summarize this podcast", "what's this episode about?"
• "transcribe this video/podcast"
• "donne-moi un résumé de cette vidéo/podcast/article"

Content type detection

Identify the content type from the URL or user description:

Type	URL patterns / clues
Text / Article	Any URL not matching video/podcast patterns, or raw text pasted by user
YouTube Video	youtube.com/watch, youtu.be/, youtube.com/shorts/
Podcast	Spotify, Apple Podcasts, podcast RSS feeds, or user explicitly says "podcast"

---

1. Text / Article

Goal: Extract the readable text from a web page and summarize it.

Steps:

1. Use web_fetch with the URL to get the page content.
2. If the content is too long (> 30,000 chars), work with the first 30,000 chars.
3. Produce a structured summary:
   • Title of the article
   • Key points (3–7 bullet points)
   • Summary (1–2 paragraphs)

Example flow:

web_fetch(url: "https://example.com/article")
→ Read the returned text
→ Generate the summary from the extracted content

If web_fetch fails (e.g. paywall, bot protection), tell the user and suggest they paste the text directly.

---

2. YouTube Video

Goal: Get the video transcript/subtitles and summarize them.

Steps:

1. Extract the video ID from the URL.
   • youtube.com/watch?v=VIDEO_ID → extract VIDEO_ID
   • youtu.be/VIDEO_ID → extract VIDEO_ID
2. Fetch the transcript using the public YouTube transcript endpoint via run_terminal_command:

curl -s "https://www.youtube.com/watch?v=VIDEO_ID" | grep -o '"captionTracks":\[.*?\]' | head -1

3. If that gives you a baseUrl, fetch the XML captions:

curl -s "CAPTION_BASE_URL" | sed 's/<[^>]*>//g'

4. Alternative — more reliable method: Use yt-dlp (if installed) to extract subtitles:

yt-dlp --skip-download --write-auto-sub --sub-lang en,fr --sub-format vtt --convert-subs srt -o "/tmp/yt_%(id)s" "VIDEO_URL" 2>/dev/null && cat /tmp/yt_*.srt

5. Alternative — fallback via web: If other methods fail, try fetching a transcript from a third-party service:

curl -s "https://youtubetranscript.com/?server_vid2=VIDEO_ID"

6. Once you have the transcript text, produce a structured summary:
   • Video title (extract from the page if possible)
   • Key points (3–7 bullet points)
   • Summary (1–2 paragraphs)
   • If the user asked for the full transcript, return it formatted cleanly.

Important: If no subtitles/captions are available (some videos have none), tell the user clearly: "This video has no available subtitles or auto-generated captions. I cannot transcribe it."

---

3. Podcast

Goal: Get a transcript or show notes and summarize them.

Steps:

1. Try to find a transcript:

   • Use web_search to find a transcript: "PODCAST_NAME EPISODE_TITLE transcript"
   • Some podcast platforms provide transcripts directly (e.g., Spotify shows, some Apple Podcast pages).

2. If a transcript page is found, use web_fetch to extract the text and summarize it.

3. If no transcript is available, try to get show notes:

   • Use web_fetch on the podcast episode page to extract the description/show notes.
   • Summarize from the show notes, but tell the user it's based on show notes, not a full transcript.

4. If the podcast has an RSS feed URL, parse it:

curl -s "RSS_FEED_URL" | grep -oP '<description><!\[CDATA\[.*?\]\]></description>' | head -5

5. Produce a structured summary:
   • Podcast name and episode title
   • Key topics (3–7 bullet points)
   • Summary (1–2 paragraphs)
   • Source: indicate whether the summary is from a transcript, show notes, or description.

Important: If no transcript or meaningful show notes can be found, tell the user clearly and suggest they provide a direct link to a transcript page or the audio file if they have a transcription tool.

---

Output format

Always structure your summary clearly:

## 📝 Summary: [Title]

**Source:** [Article / YouTube Video / Podcast]
**URL:** [original URL]

### Key Points
- Point 1
- Point 2
- ...

### Summary
[1–2 paragraph summary]

Language

• Detect the language of the source content and write the summary in the same language.
• If the user asks in French, answer in French regardless of source language.
• If the user explicitly asks for a different language, use that.

Tips

• Always try web_fetch first for articles — it's the fastest path.
• For YouTube, try the yt-dlp method first if available, fall back to curl scraping.
• For podcasts, search for transcripts before resorting to show notes.
• If the content is very long, summarize in chunks and then synthesize.
• If the user says "transcribe" instead of "summarize", return the full transcript (cleaned up) rather than a summary.