repurpose-talk v1.1.0A

Immutable. This exact content never changes and is served at /api/v1/blob/87d390c8ffd8b469.

---
name: repurpose-talk
description: Use when the user just gave a talk, panel, fireside chat, keynote, workshop, podcast, or webinar and wants content made from it, or shares an article (their own published piece or someone else's) to break into posts, carousels, and short-video scripts — triggers like "I just gave a talk", "the panel is done", "just finished speaking", "turn my talk into LinkedIn posts", "repurpose this recording/transcript into posts, notes, or clips", "break this article into social posts", or /repurpose-talk. Works from a recording, transcript, slides, a from-memory brain-dump, or an article URL or file.
trigger: /repurpose-talk
version: 1.1.0
---

# Repurpose Talk

Turn a single speaking engagement into maximum content leverage. Run this after any talk, panel, fireside chat, or workshop.

## Trigger

`/repurpose-talk` or when the user says "I just gave a talk", "the panel is done", "just finished speaking", or shares an article and asks to turn it into content

## Input

Ask for (in this order, skip any they don't have):
1. **Recording** - video/audio file path or link
2. **Transcript** - if available (check meeting note tools, Google Docs, or manual notes)
3. **Slides** - if used
4. **If none of the above**: ask them to brain-dump the key points, best audience reactions, and any quotes they remember
5. **Article** (Article mode): a URL or file of an article. Read it in full with WebFetch. If the page blocks fetching, ask the user to paste the text. Never work from the headline or a summary alone.

### Article mode

When the input is an article instead of a talk, run the same phases with these changes:

1. **Ask whose article it is** before extracting anything.
   - **The user's own published piece:** posts can quote it freely, and the link goes in the first comment, not the post body.
   - **Someone else's article:** every piece names the source and adds the user's own take or experience. Never restate the article as if it were the user's idea, and never quote more than one or two sentences from it.
2. **Phase 1 extracts, for an article:** the thesis in one sentence, 3-5 claims with their data (keep the original source of every number), 1-2 examples or cases, and the questions the article leaves open.
3. **Find the user's angle for each claim.** Search the vault (journals, meeting notes, project notes) for something the user lived that connects to the claim: a result, a mistake, a client story. If nothing connects, ask the user for one. A piece with no personal angle is a summary, and summaries of someone else's article underperform. Drop claims that end up with no angle rather than publishing them as summaries.
4. **Phase 1.5** uses the same virality categories, applied to claims instead of clips.
5. **Phase 2 package for an article:** 3-5 LinkedIn posts, 1-2 carousels (one slide per idea, 6-10 slides, title slide states the claim), 1-3 short-video scripts (hook in the first line, one idea, under 60 seconds read aloud), and 5-8 short-form notes. Skip the Video Clip Plan and the Article Seed.
6. **Output** for an article: `Article Repurpose - [Article Title] - YYYY-MM-DD.md` in the Writing folder.

## Process

### Phase 1: Extract (from whatever input is available)

Pull out:
- **3-5 key insights** they shared (the things that made people nod or take notes)
- **1-2 stories** they told (personal anecdotes are the highest-performing content)
- **Best one-liners** (quotable sentences, these become standalone posts)
- **Audience questions** that came up (each question is a content seed)
- **The main thesis** of the talk in one sentence

### Phase 1.5: Score clip candidates against the virality framework

For each candidate moment (one-liner, story beat, question, insight), tag which virality category it hits. Cherry-picked from SamurAIGPT/Generative-Media-Skills `ai-clipping` SKILL.md (MIT) — 8 categories that explain WHY short-form clips travel:

- **Hook moments** — strong opening line that stops the scroll
- **Emotional peaks** — laughter, anger, vulnerability, awe
- **Opinion bombs** — spicy, contrarian, debate-bait takes
- **Revelation moments** — "wait, what?" reframes
- **Conflict** — disagreement, tension, callouts
- **Quotable lines** — tight, screenshot-worthy phrasing
- **Story peaks** — climax of a narrative arc
- **Practical value** — actionable insight a viewer will save

A clip that hits 2+ categories is a stronger pick than one that hits 1. A clip that hits 0 isn't worth packaging — drop it.

The taxonomy is FOR RANKING + EXPLAINING, not for inventing content. Every clip still has to come from the actual talk; never confabulate a "spicy take" the speaker didn't make.

### Phase 2: Generate Content Package

Create all of the following:

**LinkedIn Posts (3-5)**
- Each post: 150-300 words, first-person, starts with a hook line
- One post per key insight
- No AI-sounding language. Run through humanizer principles if available.
- End each with a question to drive comments
- Format: hook line, story/context, insight, question

**Short-Form Notes (5-8)**
- Short-form (1-5 sentences each)
- Mix of: one-liners, micro-stories, questions, frameworks
- If using Substack: append to the Substack Notes drafts file
- If using another platform: save to a drafts file in the vault

**Article Seed (1)**
- If the talk contains enough for a full essay, create an article seed
- Title + 2-3 paragraph draft expanding the talk's core thesis

**Video Clip Plan (if recording exists)**
- Timestamp markers for the 3-5 best 30-60 second clips
- Suggested caption for each clip
- Which platform each clip is best for (LinkedIn, Instagram Reels, TikTok)

### Phase 3: Schedule

- Suggest a 2-week posting calendar: which piece goes out when
- Rule: never post more than 1 LinkedIn post per day, 2 short-form notes per day
- Space the content to maximize reach (don't dump everything day 1)

### Phase 4: Cross-pollinate

- Identify any business angles (if the talk was about their industry/product)
- Flag any investor-relevant soundbites for fundraising narrative
- Note any audience members worth following up with (add to CRM if names given)

## Output

Save the full content package to the Writing folder: `Talk Repurpose - [Event Name] - YYYY-MM-DD.md`
Append short-form notes directly to the drafts file.
Open the main file in Obsidian when done (if Local REST API is available).

## Rules

- Use the user's actual voice. If a voice-profile skill or note exists in the vault, load it first; otherwise read 2-3 of their published posts to calibrate tone.
- No em dashes anywhere.
- Run humanizer principles on every piece if the humanizer skill is available.
- If the talk was in a non-English language, create content in BOTH languages (primary + secondary audience).
- Every LinkedIn post must be standalone (someone who didn't attend the talk should understand it).