j-star-films-studios High Fidelity Data Extraction Protocol
This skill enables the agent to extract deep intelligence from social media platforms (Instagram, TikTok, YouTube, etc.) and complex web environments with high precision.
🚀 Core Philosophy: The Extraction Spectrum
When tasked with extraction, always check the prompt for specific limitations. If no limitations are provided, assume Standard Level.
Extraction Capability Matrix
| Feature | Basic (Standard) | Deep (Advanced) | Elite (Full Intelligence) |
|---|---|---|---|
| Captions | Full Text + Hashtags | + Edited timestamps | + OCR from video overlays |
| Comments | Top 3-5 (Top Level) | Top 10 + Threaded replies | Full sentiment & pain point mapping |
| Engagement | Likes / Views count | Engagement Rate % | Share/Save estimates & Viral Velocity |
| Brand Intel | @Mentions in caption | + Link-in-bio analysis | + Competitor comparison vs Meta Ad Library |
| Visuals | Profile Screen/Grid | Individual Post Screenshots | UI/UX Reverse Engineering of Funnel |
| Technical | URL collection | Precise ISO Timestamps | API Pattern Mapping & DB Schema generation |
📊 Tabular Output Format
Unless the user specifies otherwise, all extracted data should be compiled into a Markdown table for maximum scannability.
Standard Template:
| Post Type | Caption Snippet | Top Comments (Synthesized) | Key Metrics | Brand/Tags |
| :--- | :--- | :--- | :--- | :--- |
| Pinned | "How to be..." | Users asking about discipline tips | 30k Likes | @Adidas, #NYC |
| Recent 1 | "Day in life..." | High praise for work/life balance | 15k Views | @CeraVe |
🧠 Smart Filtering Strategy (Instagram Reels)
When the goal involves "high-performing" content or maximizing engagement:
- Reels First: Navigate to the
/reels/tab immediately. View counts are not visible on the main grid but are overlayed on Reel thumbnails. - Baseline Calculation:
- Extract view counts from the first 6-12 visible Reels.
- Calculate the average (mean) view count.
- Filtration:
- Only click/extract posts that exceed this average.
- This ensures we focus valuable browser resources on proven content.
🛠️ Execution Instructions for the Subagent
- DOM First: Never navigate blindly. Use
browser_get_domto identify the obfuscated class names for captions and comments. - Surgical Navigation: To avoid 429 errors (Too Many Requests), navigate directly to post URLs once the links are gathered from the grid, rather than infinite scrolling.
- Expansion Logic: Always look for
... moreor "View all comments" buttons and trigger them via JavaScript to ensure data completeness. - Verification: Always capture a final screenshot of the "Main Extraction Target" to verify text accuracy against the visual truth.
🛑 Limitations & Compliance
- Obey Prompt Bounds: If the user says "Only get usernames," do NOT extract captions.
- Privacy: Respect platform TOS by focusing on public-facing data and ignoring private user details.
- Anti-Spam: Filter out "Promote on..." or generic bot comments automatically.