prompt_chain.md

gistfile1.txt

# GOAL
Main Goal of all of this: **Finding how to create content that resonates to build the audience faster.**

How we do this:

Analysing competitor content, finding what content works in terms of ideas, strucutres, and emotions. And then crearting a content strategy around that.

## Setup Context

Folder structure:

```
/competitor-analysis
  /input
    competitor_a.csv
    competitor_b.csv
    competitor_c.csv
  /output
    combined_posts.csv
    tagged_posts.csv
    idea_mapping.json
    tagged_posts_normalized.csv
    summary_tables.csv
    insights.md
```

Each input CSV needs:
- `post_content` (the text)
- `engagement` (likes, views, whatever you have)
- `competitor` (name or handle, or infer from filename)

---

## Prompt 1: Setup & Combine Files

```
I have a folder called /input with multiple CSV files. Each CSV contains short-form social media posts from a competitor.

Each CSV has these columns:
- post_content: the text of the post
- engagement: a number representing likes/views
- competitor: name of the competitor (or infer from filename if not present)

Your job:
1. Read all CSV files from /input
2. Combine them into one unified dataset
3. Add a "source_file" column so I know where each row came from
4. Show me a count of how many posts per competitor we're working with
5. Save the combined file as /output/combined_posts.csv

Don't start analysis yet. Just prepare the data and confirm what we're working with.
```

---

## Prompt 2: Tag Each Post (Core Idea Extraction)

```
Read /output/combined_posts.csv

For each row, analyze the post_content and add these new columns:

1. core_idea
   THE MOST IMPORTANT FIELD.
   Extract the specific claim, insight, or argument the post is making.
   Write it as ONE clear sentence.
   
   Rules:
   - Be concrete and specific, not categorical
   - Capture the ACTUAL ARGUMENT, not the topic
   - Write it as a statement someone could agree or disagree with
   
   BAD examples (too generic):
   - "content strategy"
   - "growing an audience"
   - "email marketing tips"
   - "mindset advice"
   
   GOOD examples (specific arguments):
   - "you should post daily even if the quality isn't perfect"
   - "your first 100 posts are just practice, not content"
   - "niching down too early kills most creators"
   - "one viral post matters less than 100 consistent ones"
   - "your email list is the only asset you actually own"
   - "give away your best stuff for free, sell the implementation"
   - "most creators quit right before the algorithm rewards them"
   - "repurposing is how pros create 10x more without 10x effort"

2. topic_category
   Broad bucket for grouping. Pick ONE:
   - content_creation
   - audience_growth
   - email_marketing
   - monetization
   - offer_creation
   - productivity
   - mindset
   - personal_brand
   - platform_tactics
   - other

3. hook_type
   How does the post open? Pick ONE:
   - question: asks the reader something
   - bold_claim: states something provocative or contrarian
   - story: starts with personal narrative, uses "I"
   - pain_point: calls out a frustration or struggle
   - how_to: promises a method or steps
   - stat_proof: leads with a number or result
   - curiosity_gap: teases without revealing

4. emotion
   What feeling does this trigger? Pick ONE:
   - awe: inspiring, mind-expanding, "wow I never thought of it that way"
   - anger: frustrating, calls out BS, "this is what's wrong with..."
   - anxiety: fear-based, FOMO, "you're falling behind"
   - excitement: opportunity, possibility, "imagine if you could..."
   - amusement: funny, entertaining, clever
   - validation: "you're not alone", "it's okay that you..."
   - neutral: purely informational, no emotional charge

5. share_driver
   Why would someone repost this? Pick ONE:
   - social_currency: makes sharer look smart or in-the-know
   - practical_value: genuinely useful, helps others
   - strong_opinion: polarizing, people want to agree/disagree publicly
   - relatability: "this is so me" content
   - story: narrative that pulls people in

Rules:
- The core_idea is the priority. Spend most effort here.
- Pick the DOMINANT tag for other fields, not all that apply
- When torn between two, pick the stronger one

Save the result as /output/tagged_posts.csv

After saving, show me:
- Total posts processed
- 10 sample core_ideas so I can check if you're being specific enough
- Distribution of topic_category
- Distribution of emotion
```

---

## Prompt 3: Normalize Core Ideas

```
Read /output/tagged_posts.csv

Extract all unique values from the "core_idea" column.

Many of these will express the same underlying argument in different words.

For example, these are all the SAME idea:
- "post every day even if it sucks"
- "quantity beats quality when you're starting"
- "your first 100 posts are just practice"
- "perfectionism kills creators, just ship it"

Your job:
1. Group core_ideas that make essentially the same argument
2. For each group, write ONE clear master statement that captures the insight
3. Create a mapping

Save as /output/idea_mapping.json in this format:
{
  "post daily, quality comes with reps": [
    "post every day even if it sucks",
    "quantity beats quality when starting",
    "your first 100 posts are practice",
    "perfectionism kills creators"
  ],
  "your email list is your only real asset": [
    "email list beats social followers",
    "you don't own your social audience",
    "build your list before anything else"
  ]
}

Guidelines:
- The master statement should be clear enough that someone could agree or disagree with it
- Aim for 30-60 distinct ideas (depends on dataset size)
- If an idea only appeared once and doesn't fit anywhere, keep it as its own entry
- Don't over-consolidate — "post daily" and "email beats social" are different ideas even if both are about growth

After creating the mapping, show me:
- How many unique raw core_ideas you found
- How many master ideas you consolidated them into
- The top 20 master ideas by frequency (how many posts express each)
```

---

## Prompt 4: Apply Normalized Ideas

```
Read /output/tagged_posts.csv and /output/idea_mapping.json

Replace each "core_idea" value with its master statement from the mapping.

Add a new column called "original_idea" that preserves the original extraction.

Save as /output/tagged_posts_normalized.csv

Show me:
- Confirmation it's done
- Top 25 normalized ideas by post count
```

---

## Prompt 5: Build Summary Tables

```
Read /output/tagged_posts_normalized.csv

Create these summary tables:

TABLE 1: IDEA PERFORMANCE (most important)
For each core_idea:
- post_count: how many posts express this idea
- avg_engagement: mean engagement
- max_engagement: best performing post with this idea
- top_competitor: who posts this idea most
- common_hook: most frequent hook_type used
- common_emotion: most frequent emotion used

Sort by avg_engagement descending.
Only include ideas with 3+ posts.

TABLE 2: IDEA BY COMPETITOR
Matrix showing which competitors use which ideas:
- Rows: top 30 ideas by frequency
- Columns: each competitor
- Cells: count of posts + avg engagement

This shows who "owns" which ideas.

TABLE 3: TOPIC CATEGORY PERFORMANCE
For each topic_category:
- post_count
- avg_engagement
- top_idea: best performing idea in this category
- worst_idea: lowest performing idea in this category (min 3 posts)

TABLE 4: HOOK TYPE PERFORMANCE
For each hook_type:
- post_count
- avg_engagement
- best_idea_pairing: which idea works best with this hook
- best_emotion_pairing: which emotion works best with this hook

TABLE 5: EMOTION PERFORMANCE
For each emotion:
- post_count
- avg_engagement
- top_ideas: top 3 ideas that use this emotion
- top_hooks: which hooks pair best

TABLE 6: SHARE DRIVER PERFORMANCE
For each share_driver:
- post_count
- avg_engagement
- top_ideas: top 3 ideas with this driver

TABLE 7: WINNING COMBINATIONS
Group by: core_idea + hook_type + emotion
- Only combinations with 3+ posts
- avg_engagement
- Sort descending
- Show top 30

TABLE 8: COMPETITOR OVERVIEW
For each competitor:
- total_posts
- avg_engagement
- signature_ideas: top 3 ideas they use most
- best_performing_idea: their highest engagement idea
- dominant_hook
- dominant_emotion

Save all tables to /output/summary_tables.csv

Display TABLE 1 (Idea Performance) and TABLE 8 (Competitor Overview) in full so I can review.
```

---

## Prompt 6: Generate Insights Report

```
Read /output/summary_tables.csv and /output/tagged_posts_normalized.csv

You are a content strategist analyzing what specific IDEAS resonate in this niche.

Write a clear, actionable report:

## WINNING IDEAS
List the top 10 highest-performing ideas by engagement.
For each one:
- State the idea clearly
- Show the numbers (post count, avg engagement)
- Explain WHY you think this resonates (based on emotion, hook patterns, etc.)
- Give an example of how I could express this idea in my own voice

## LOSING IDEAS
List 5-10 ideas that get posted frequently but underperform.
- Why might these be falling flat?
- Are they too generic? Wrong emotion? Overplayed?
- Should I avoid entirely or is there a reframe?

## HOOK + IDEA PATTERNS
Which hooks work best for which types of ideas?
Be specific:
- "Contrarian ideas perform 2x better with bold_claim hooks than with how_to"
- "Mindset ideas need story hooks to land, stat_proof kills them"

## EMOTION PATTERNS
Which emotions drive engagement in this niche?
- Is it anger? Validation? Awe?
- Any ideas that only work with a specific emotion?

## COMPETITOR POSITIONING
For each competitor:
- What ideas do they "own"? (post most + perform well)
- What's their emotional signature?
- What's working for them that others aren't doing?

## GAPS AND OPPORTUNITIES
Based on everything:
- Which winning ideas are underserved? (work well but few posts)
- Which idea + hook + emotion combos are untapped?
- Where's the white space I could own?

## MY CONTENT PLAN: 15 POST IDEAS
Based on the data, give me 15 specific posts I should create.

Format each as:
IDEA: [the core argument I'm making]
HOOK: [how I'll open it]
EMOTION: [what feeling I'm targeting]
ANGLE: [2-3 sentences describing the specific take and why it's differentiated]
DATA BACKING: [why this should work based on the analysis]

Make these specific enough that I could write the post tomorrow.

Save as /output/insights.md
```

---

## Prompt 7: Competitor Deep-Dive (Optional)

```
Read /output/tagged_posts_normalized.csv

Filter to only posts from: {{competitor_name}}

Analyze their content strategy:

## THEIR TOP 10 POSTS
List their 10 highest-engagement posts.
For each show:
- The core idea
- Hook type
- Emotion
- Why you think it worked

## THEIR BOTTOM 10 POSTS
List their 10 lowest-engagement posts (that have at least average effort).
For each show:
- The core idea
- Hook type  
- Emotion
- Why you think it flopped

## THEIR SIGNATURE IDEAS
What ideas do they return to repeatedly?
Which of these actually perform vs which are "pet topics" that flop?

## THEIR FORMULA
Based on winners vs losers:
- What hook types work for them?
- What emotions work for them?
- What topics should they lean into?
- What should they stop posting about?

## WHAT I CAN STEAL
5 specific, concrete things I can take from their winning posts:
- Specific ideas worth exploring
- Hook structures that work
- Emotional angles that land
- Phrases or framings worth adapting

Save as /output/deep_dive_{{competitor_name}}.md
```

---

## Prompt 8: Cross-Competitor Comparison (Optional)

```
Read /output/tagged_posts_normalized.csv

Compare: {{competitor_a}}, {{competitor_b}}, {{competitor_c}}

## IDEA OVERLAP
Which ideas do ALL of them post about?
How does performance compare when they post about the same idea?
(e.g., "When Competitor A posts about 'email > social', avg engagement is 1,200. When B posts the same idea, it's 600. Why?")

## UNIQUE TERRITORY
Which ideas does each competitor "own" that others don't touch?
Are these working? Untapped opportunity or dead end?

## STYLE COMPARISON
| Competitor | Dominant Hook | Dominant Emotion | Avg Engagement |
Compare their approaches.

## WHO'S WINNING WHICH IDEAS
For the top 20 ideas overall, who performs best on each?
| Idea | Comp A Avg | Comp B Avg | Comp C Avg | Winner |

## THE PLAYBOOK I'D STEAL FROM EACH
For each competitor, what ONE thing would I take?
- Competitor A: [specific tactic]
- Competitor B: [specific tactic]
- Competitor C: [specific tactic]

## WHITE SPACE
Based on all three:
- What ideas work but nobody's doubling down on?
- What emotional angles are underused?
- What's my positioning opportunity?

Save as /output/competitor_comparison.md
```

---

## Prompt 9: Idea-Specific Deep Dive (Optional)

If a particular idea interests you and you want to see exactly how it's being executed:

```
Read /output/tagged_posts_normalized.csv

Filter to posts where core_idea = "{{specific idea}}"

Show me:

## ALL POSTS WITH THIS IDEA
List every post that expresses this idea.
Include:
- Competitor
- Full post text
- Engagement
- Hook type
- Emotion

Sort by engagement descending.

## WHAT SEPARATES WINNERS FROM LOSERS
Compare the top 5 vs bottom 5 posts with this idea:
- How do they open differently?
- What's the tone difference?
- Length difference?
- Any specific phrases in winners?

## HOW I SHOULD EXECUTE THIS IDEA
Based on what works:
- Best hook type for this idea
- Best emotion to pair with it
- Specific angle or framing that performs
- What to avoid

Write me 3 different versions of this idea I could post, each with a different hook style.

Save as /output/idea_deep_dive_{{idea_slug}}.md
```

---

That's the complete revised sequence. The key difference is you're now tracking **specific arguments** rather than generic categories, so your final output tells you exactly which ideas to use, not just which topics to cover.