content-reflect

name	content-reflect
description	User wants to analyze their own LinkedIn posts (later YouTube) to understand what's working, what's not, and what to pos

/content-reflect — Own Content Performance Analysis

When to use

User wants to analyze their own LinkedIn posts (later YouTube) to understand what's working, what's not, and what to post next.

Setup Wizard (First Run)

Check MEMORY.md for SETUP_CONTENT_REFLECT_COMPLETE: true. If missing:

S1 — Verify Data Source

Test Apify: Scrape user's own LinkedIn activity page
  Pass: Returns post list with engagement data
  Fail: Fall back to DOM extraction

Test DOM fallback: Navigate to linkedin.com/in/[user-slug]/recent-activity/all/
  Pass: Activity page loads, posts visible
  Fail: "Cannot access LinkedIn — log in first"

S2 — Get User's LinkedIn Profile

Read MEMORY.md for LinkedIn URL
If missing: ask "What's your LinkedIn profile URL?"
Extract slug for activity page navigation

S3 — Create Analysis File

Create if missing: memory/content-analysis.md (with header template)
Write SETUP_CONTENT_REFLECT_COMPLETE: true to MEMORY.md

---

Instructions

Step 1: Scrape Own Posts (READ — Apify preferred, DOM fallback)

With Apify:

Apify scrape: user's LinkedIn activity page
Returns per post: text preview, post date, likes, comments, shares, impressions (if available)
Pull last 10 posts minimum

With DOM fallback:

Navigate to linkedin.com/in/[slug]/recent-activity/all/
JS extract from each post in feed:
  - Post text (first 200 chars)
  - Reaction count (span with reaction count)
  - Comment count
  - Repost count
  - Post date
Scroll to load 10 posts, extract each

DOM extraction pattern:

const posts = document.querySelectorAll('.profile-creator-shared-feed-update__container');
posts.forEach(post => {
  const text = post.querySelector('.feed-shared-update-v2__description')?.textContent;
  const reactions = post.querySelector('.social-details-social-counts__reactions-count')?.textContent;
  const comments = post.querySelector('.social-details-social-counts__comments')?.textContent;
});

Step 2: Analyze Patterns

For each post, extract:

Quantitative:

• Engagement rate: (likes + comments × 3 + shares × 2) / estimated reach
• Comment-to-like ratio (high ratio = discussion driver, low = passive consumption)
• Best performing post (absolute engagement)

Qualitative:

• Hook analysis: what was the first line? (question, stat, contrarian take, story)
• Format: text-only vs image vs video vs carousel
• Length: short (<100 words) vs medium (100-300) vs long (300+)
• Topic category: build log, opinion, tool review, personal story, industry take
• Time of posting: morning/afternoon/evening, day of week
• CTA: open question, link, none

Step 3: Pattern Recognition

Compare top 3 vs bottom 3 posts:

What do the winners have in common?
  - Hook type
  - Format
  - Length
  - Topic
  - Time posted

What do the losers have in common?
  - Same analysis

Step 4: Write Analysis

Update memory/content-analysis.md:

# Content Performance Analysis
Last updated: [date]
Posts analyzed: [count]
Period: [date range]

## Top 3 Posts
| # | Date | Hook (first line) | Format | Likes | Comments | Why it worked |
...

## Bottom 3 Posts
| # | Date | Hook (first line) | Format | Likes | Comments | Why it underperformed |
...

## Patterns
- Best hook type: [X]
- Best format: [X]
- Best length: [X]
- Best posting time: [X]
- Topic that resonates most: [X]

## Recommendations for Next 3 Posts
1. [specific recommendation tied to a pattern]
2. [specific recommendation]
3. [specific recommendation]

Step 5: YouTube Extension (Future)

When YouTube channel is live, add:

Apify scrape own last 5 videos
  → CTR, watch time, retention curve
  → Title + thumbnail analysis
  → Hook analysis (first 30 seconds)
  → Save to memory/youtube-analysis.md

---

MCP Tools Used

• Apify MCP (READ — scrape own activity page) OR claude-in-chrome (DOM fallback)
• No WRITE operations needed — analysis only