SKILL FILE

Instagram Channel Deep Dive with AI

Scrape any Instagram profile's last 25 reels, transcribe each one with Relevance AI or Apify, and turn them into a strategic content audit + queryable corpus you can chat with.

25 reels scraped + transcribed per channel
10 min from URL to strategic audit
~$0.90 per channel (cached re-runs free)
Download Skill File ↓

How channel deep-dives flow into your content engine

One scrape becomes a queryable corpus that powers every department

/ig-channel @handle 25 reels scraped + transcribed + clustered
1 Playwright scrapes profile
2 Relevance AI / Apify transcribes
3 Claude clusters the corpus
4 Queryable archive on disk
Marketing
  • Hooks library across N creators (ranked by recency × engagement)
  • Voice-cloned scripts in any creator's exact cadence
  • Format patterns to copy verbatim — length, shot, on-screen text, CTA
  • Repurposement seeds: 'steal-this' list of derivative content ideas
Sales
  • Lead-magnet pattern mining — every 'Comment X to get Y' CTA across creators
  • Find creators talking about your ICP problems → outbound list
  • Competitor positioning intel — how they frame the problem they solve
  • Skill-stack research: what tools and frameworks they teach
Growth
  • Niche onboarding — reverse-engineer 5 top creators in a single afternoon
  • Content calendar planning — steal the topic mix that's already working
  • Cross-creator topic frequency → trend signal
  • Posting cadence benchmarks (how often, what days, what beats)
Voice-Cloned Script
Hooks Library
Lead-Magnet Pattern Map
Niche Content Calendar
Events Tracked
Every claim cites a real reel ID
Transcripts cached → corpus compounds
/ig-channel @handle 25 reels scraped + transcribed + clustered
1
Playwright scrapes profile Real Chrome session, 25 reel URLs harvested in 4 scrolls
2
Relevance AI / Apify transcribes Full transcript + caption + hashtags + engagement per reel
3
Claude clusters the corpus Pillars, hooks, formats, what's working, what to steal
4
Queryable archive on disk Every transcript persists — chat with it anytime
Marketing
  • Hooks library across N creators (ranked by recency × engagement)
  • Voice-cloned scripts in any creator's exact cadence
  • Format patterns to copy verbatim — length, shot, on-screen text, CTA
  • Repurposement seeds: 'steal-this' list of derivative content ideas
Sales
  • Lead-magnet pattern mining — every 'Comment X to get Y' CTA across creators
  • Find creators talking about your ICP problems → outbound list
  • Competitor positioning intel — how they frame the problem they solve
  • Skill-stack research: what tools and frameworks they teach
Growth
  • Niche onboarding — reverse-engineer 5 top creators in a single afternoon
  • Content calendar planning — steal the topic mix that's already working
  • Cross-creator topic frequency → trend signal
  • Posting cadence benchmarks (how often, what days, what beats)
Content Outputs
Voice-Cloned Script from marketing
Hooks Library from marketing
Lead-Magnet Pattern Map from sales
Niche Content Calendar from growth
Everything Tracked
Every claim cites a real reel ID
Transcripts cached → corpus compounds

What this skill file teaches Claude

Drop one markdown file into your repo. Claude Code learns how to run this entire workflow.

1

Profile-Level Reel Scraping

Playwright drives a real Chrome session against instagram.com/{handle}/reels/ — not headless, not detectable. Harvests the last 25 reel URLs in one pass.

2

Transcript Engine — Two Paths

Clone our Relevance AI tool with one click, or skip Relevance entirely and call Apify's instagram-reel-scraper directly from a Node helper. Same output, your choice.

3

Disk-Cached Transcripts

Every reel response is saved to channels/{handle}/transcripts/{reel_id}.json. Re-running on the same channel costs $0 — your private creator-knowledge corpus compounds.

4

Strategic Channel Audit

Claude clusters 25 reels into content pillars (% share + reel IDs that prove each one), recurring hooks (with frequency counts), format patterns, and a steal-this list.

5

Queryable Corpus

Output isn't just a static report — Claude has every transcript on disk. Ask any question and get answers grounded in real reel IDs (RAG without vector DB plumbing).

6

HTML Report

Auto-generates channels/{handle}/report-YYYY-MM-DD.html — a polished one-page audit with stats, pillars, hooks, and per-reel breakdowns. Shareable link.

Build it with plain English

Tell Claude Code what to do. It handles the rest.

claude — solostack/
you: |
Scraping @gregisenberg — target 25 reels...
✓ Profile loaded, scrolling to collect reels
✓ Collected 25 reels in 4 scrolls
✓ Transcribing reels via Relevance AI (sequential)
  [01/25] DXxfIc6zC4G ✓
  [02/25] DXx4_LCMO3p ✓
  ... 23/25 transcribed (2 had no audio)
✓ Saved 25 transcript files to channels/gregisenberg/transcripts/
✓ Clustering content pillars... 5 found
✓ Extracting recurring hooks... 14 patterns
✓ Writing audit report

Top pillars: Community-led growth (~40%), Startup ideas (~25%), AI tools (~20%)
Report: channels/gregisenberg/report-2026-05-04.html
you: |
Greg's #1 hook (used in 11/25 reels):
→ Curiosity gap → contrarian claim → number

Examples:
• DXxfIc6zC4G — 'The internet isn't a feed anymore, it's a stack of communities'
• DX5SKOpD3CW — 'Here's a $1M startup idea you can build with AI today'
• DX38njeJh_G — 'I'd build this if I was starting over in 2026'

Pattern: opens with a strong claim, immediately follows with a specific number or framework. Top performers add a $-figure in the first 3 seconds.
you: |
Hook: 'Most founders are about to spend $542/mo on SaaS they don't need.'

[Beat 1 — Problem agitate, 0-15s]
If you're running a $0-1M business and you've already got HubSpot, Mailchimp, ContentCal, Calendly — your data lives in 4 dashboards and nothing talks to each other.

[Beat 2 — Education, 15-50s]
Here's the actual cost: not the $542. The cost is that lead → contact → customer is three rows in three databases.

[Beat 3 — Soft pivot, 50-90s]
I replaced 12 SaaS tools with one repo. One CRM record, every stage. Comment 'STACK' and I'll send the open-source repo.

— Voice notes: Greg uses short clipped sentences, 1-line beats, signature phrase 'here's the thing'. 90s talking-head + screen demo at minute 1.

What you can build with this

Get the full skill file

Everything above is 80% of the skill file. Download the complete version with full implementation details, agent prompts, and ready-to-run scripts.

Common questions

No. The skill ships with two transcript engines: Option A clones our Relevance AI tool with one click, Option B skips Relevance entirely and calls Apify's instagram-reel-scraper directly from a tiny Node helper. Both return the same shape.
Playwright connects via CDP to a real Chrome session running on a dedicated profile (~/.claude-browser). No headless flag, no automation banner, no navigator.webdriver. Instagram can't distinguish it from a logged-in human.
No vector DB. Claude has all 25 transcripts on disk as JSON files. When you ask a question, Claude reads the analysis JSON + relevant transcripts and answers grounded in real reel IDs. Same effect as RAG, zero plumbing.
About $0.90 — that's 25 reels × ~3.66 Relevance credits each. Re-runs on the same channel are free because transcripts are cached on disk. A 10-creator knowledge base costs ~$9 total and compounds every time you add a new creator.
Yes for the scraping, no for the transcription. Reel transcription must run sequentially per channel — parallel calls trigger 504 timeouts on both Apify and Relevance. The skill is designed for one channel at a time.
The transcript engine returns a stub with `error: 'No transcript available'`. The skill saves the stub to disk so it's not retried, then continues. Image-only reels usually account for 1-3 of every 25.

Ready to automate?

SoloStack gives you every skill pre-installed — scraping, marketing, sales, CRM, and more. One repo. Every department.

Book a Call →