mcpbeat Sign in

Transcript Intelligence Agent Skill

Use when the user wants to summarize, analyze, or repurpose transcripts from TikTok, Instagram, YouTube, Facebook, X/Twitter, LinkedIn, Rumble, or Reddit video posts. Extracts hooks, claims, quotes, content atoms, themes, and reusable scripts.

942 tokens
context cost
the whole folder, loaded on every use
1
files
instructions only
0
copies elsewhere
how many repositories repackaged it
1780
stars on the repo
on the repository, not the skill itself

Install

one command, takes just this skill from the repository
npx skills add https://github.com/ScrapeCreators/social-media-research-skills --skill transcript-intelligence

What it tells the agent to use

found in the instruction text
Bash runs shell commands — read the instruction before connecting
WebFetch fetches pages from the network

The instruction itself

7 sections, as written by the author

Transcript Intelligence

Overview

Turn public video transcripts into useful research and content assets. This skill is for extracting signal from spoken social video: hooks, claims, stories, objections, examples, CTAs, and reusable content angles.

When to Use

Use this skill when the user asks to:

  • summarize a video, reel, short, TikTok, podcast clip, or social video
  • analyze a creator's hooks or speaking style
  • extract quotes, claims, examples, and CTAs from transcripts
  • turn transcripts into social posts, scripts, newsletters, or content ideas
  • compare what multiple creators say about a topic

Transcript Sources

| Platform | Endpoint |

|---|---|

| TikTok | /v1/tiktok/video/transcript |

| Instagram | /v2/instagram/media/transcript |

| YouTube | /v1/youtube/video/transcript, /v1/youtube/video |

| Facebook | /v1/facebook/post/transcript |

| X/Twitter | /v1/twitter/tweet/transcript |

| LinkedIn | /v1/linkedin/post/transcript |

| Rumble | /v1/rumble/video/transcript |

| Reddit video | /v1/reddit/post/transcript |

If a detail endpoint already includes transcript text, use it. If transcript is unavailable, say so and fall back to title/caption/description only.

Workflow

  • Collect URLs or discover videos
  • If URLs are provided, fetch each transcript directly.
  • If a creator/channel is provided, first fetch recent posts/videos, then choose relevant videos.
  • Extract transcript text
  • Preserve timestamps if provided.
  • Keep source URL with each transcript.
  • Do not hallucinate missing captions.
  • Segment the transcript

Break into:

  • hook/opening
  • setup/context
  • main claim or lesson
  • evidence/examples
  • payoff
  • CTA
  • Analyze the content

Extract:

  • exact hooks
  • claims and contrarian takes
  • stories
  • frameworks
  • objections addressed
  • emotional language
  • quotable lines
  • content atoms that stand alone
  • Synthesize across multiple transcripts
  • Cluster by topic and angle.
  • Count recurring themes.
  • Identify repeated hook formulas.
  • Flag the strongest examples with citations.

Output Format

# Transcript Intelligence Report

## TL;DR
- Main themes:
- Strongest hooks:
- Best reusable ideas:

## Transcript-by-Transcript Notes
### [Video title or URL](url)
- Hook: "..."
- Core idea: ...
- Best quote: "..."
- CTA: ...
- Content atoms:
  1. ...

## Patterns Across the Set
| Pattern | Evidence | Example URLs |
|---|---|---|

## Hooks Swipe File
- "..."
- "..."

## Repurposing Ideas
- LinkedIn post:
- X thread:
- Short-form script:
- Newsletter section:

Common Pitfalls

  • Do not summarize from the title alone if the user asked for transcript analysis. Fetch transcripts first.
  • Do not lose exact wording. Hooks and quotes are more valuable verbatim.
  • Do not treat AI-generated transcript text as perfect. If wording seems garbled, mark it as approximate.
  • Do not mix source attribution. Every quote should trace back to a URL.

How to use it

Copy the folder

Take scrapecreators/transcript-intelligence from the repository into ~/.claude/skills for personal use, or into .claude/skills inside a project.

Check the name does not clash

The agent identifies a skill by the name field in its header. Two skills with the same name cannot sit side by side — one of them will be ignored.