Stop prompting blind. Steal the cognitive structure.
The Problem
Tired of AI-generated articles that all sound the same — "as we all know", "it's worth noting", "undeniably"?
The magic of great writing was never about fancy vocabulary. It's about the underlying cognitive architecture, emotional tone, and rhythmic flow.
style-decoder is a minimalist CLI tool written in pure Rust. In 3 seconds, it穿透 any website's dynamic rendering and reverse-engineers any top-tier article into a 10-dimension System Prompt that LLMs can directly understand.
This is a physical cheat code for super-individuals in the AI era.
Core Capabilities
- Deep Fetch: Powered by Jina Reader API, bypassing complex anti-scraping mechanisms to extract clean text.
- 10-Dimension Reverse Analysis: Automatically analyzes the target article's logical skeleton, emotional tone, parallelism frequency, metaphor habits, and 7 other core dimensions.
- One-Click Weaponization: The output isn't an analysis report — it's a System Prompt ready to feed into any LLM.
- Geek-Grade Performance: Compiled Rust, single binary, zero bloated dependencies.
Quick Start
No coding knowledge required. No Rust environment setup needed:
- Download: Get the latest binary from Releases.
- Configure: Create a
.env file with your LLM API credentials:
API_KEY=sk-or-v1-your-api-key
BASE_URL=https://openrouter.ai/api/v1
MODEL=anthropic/claude-3-5-sonnet-20241022
- Run: Execute with the article URL:
./style-decoder https://mp.weixin.qq.com/s/some-viral-article
The tool outputs a deeply structured Prompt in the terminal and asks if you want to save it as a .md file.
Beyond the Tool
style-decoder is just a single probe module in my personal automation workflow.
I've built a complete AI cognitive content generation engine that supports:
- Batch processing of multiple inspiration sources (concurrent)
- CoT dual-channel rendering: logic topology outline first, then dense content filling
- Automated typesetting and clipboard injection
Links
GitHub