# The Four Caption Prompts

**From Chapter 5.2 of *The 4-Hour Side Hustle*.**

Four prompts, one per platform. Upload this file as its own Claude Project — or add it to your main Project and call the platform by name.

**Never run one prompt for all four platforms.** A single prompt produces four near-identical captions, and it fails because the platforms are different rooms. LinkedIn rewards a specific professional observation with a line break after the first sentence. TikTok rewards a hook in the first four words and punishes anything that reads like marketing. Instagram tolerates length. X does not.

In the clip chain, this is module 6, and each platform gets its own call.

---

## The shared preamble

Every one of the four starts here. Paste this, then the platform block, then the transcript.

> You are writing a social caption for a short video clip from my long-form content. My Business Brief, writing samples, and buyer vocabulary are in this project — use them. Write as me, not about me.
>
> **Input:** the clip transcript is below, along with the topic of the source video.
>
> **Hard rules:**
> - Use only claims, numbers, and examples that appear in the transcript. Invent nothing.
> - No emoji. No "In today's world." No rhetorical question as the opening line.
> - Do not describe the video ("In this clip I discuss…"). Say the thing itself.
> - If the transcript is too fragmentary to make a coherent point, reply only with `WEAK CLIP` and stop.

---

## 1. LinkedIn

> **Platform — LinkedIn:**
> - First line is a complete, specific claim that stands alone. It is the only line most people will read.
> - Then a line break, then 3–5 short paragraphs of one to two sentences.
> - Close with an observation, not a call to action. No "what do you think?"
> - 120–200 words.
> - 3 hashtags, lowercase, industry-specific, no generic tags like #business.
>
> **Transcript:**

---

## 2. TikTok

> **Platform — TikTok:**
> - Hook in the first four words. Not the first sentence — the first four words.
> - 40–70 words total.
> - Conversational to the point of being blunt.
> - No line breaks.
> - 3 hashtags: one broad, two niche.
>
> **Transcript:**

---

## 3. Instagram

> **Platform — Instagram:**
> - 80–150 words.
> - The first line must work as a standalone hook, because everything after it is truncated until someone taps "more."
> - Line breaks between thoughts.
> - 5 hashtags.
>
> **Transcript:**

---

## 4. X

> **Platform — X:**
> - Under 280 characters — or a 3-post thread, but only if the transcript genuinely supports one. If it doesn't, write the single post.
> - No hashtags at all.
>
> **Transcript:**

---

## About `WEAK CLIP`

This is the instruction worth defending hardest, and the one people delete first because it feels like giving the model permission to be lazy.

Some clips clear a 70 virality score and are still fragments that don't say anything complete. Without an escape hatch, the model will dutifully manufacture a point the transcript doesn't support — and that caption goes out under your name, attached to a video that doesn't say what the caption claims.

Giving it permission to refuse costs nothing and prevents the most common form of caption embarrassment.

**In the chain:** filter `WEAK CLIP` responses out with a router before the Buffer module. Log them rather than discarding them — a clip that produced `WEAK CLIP` on three platforms is usually a clip worth re-cutting, and that's useful information about your source video, not just about the caption.

---

## When output is wrong

Add material, not adjectives. If captions sound off, the fix is another writing sample in the Project — not another instruction in the prompt.

The one exception: if captions are consistently *too long* or *too short*, that genuinely is a prompt problem. Tighten the word count and give a hard number, not a range.

---

## Checklist

- [ ] All four platform blocks in one Project, or this file uploaded whole
- [ ] Business Brief and writing samples in the same Project
- [ ] One call per platform in the chain — never one call for four
- [ ] `WEAK CLIP` instruction present in every one of the four
- [ ] Router filtering `WEAK CLIP` before the scheduling module
- [ ] `WEAK CLIP` responses logged, not discarded
- [ ] Hashtag counts match the platform: 3 / 3 / 5 / none
- [ ] Tested against one real transcript on each platform before the chain runs unattended