Deep Thought Distillation & Style Replication

The Problem to Solve

General "de-AI-ization" tools or writing templates can make text sound human-written, but fail to solve a more fundamental problem: how to make AI-generated text sound like a specific person—not just in diction, but crucially, in their way of thinking. When the goal is to replicate an author's deep insights, unique voice, and intrinsic logic, simple style mimicry often remains superficial and misses the "soul."

How the Skill Works

This skill reverse-engineers an author's cognitive patterns and expressive system from their provided articles through a rigorous five-stage process, culminating in a high-fidelity clone.

Core Capability: Dual-Engine Deep Analysis

The core is a "dual-engine" architecture that deconstructs the author on two parallel levels:

1. Engine A: Thought Distillation (Dimension 9): Extracts the author's values, worldview, self-identity, thinking habits, and taboos from the text. The core assumption is: A person's thinking framework inevitably leaks into their writing.
2. Engine B: Writing Style Distillation (Dimensions 1-8): Performs multi-dimensional quantitative analysis of the text, covering the lexical layer, syntactic layer, tonal layer, discourse structure, rhetoric & expression, sentiment & attitude, content inclination & perspective, and format adaptation analysis.

The critical innovation lies in Phase 2.5's cross-engine verification: analyzing how thinking traits map to and drive specific writing styles, thereby extracting the most fundamental "thought DNA". For example, a thinking trait like "a default distrust of human nature" might map to a writing style where "characters are always betrayed in stories" and "direct expressions of trust are rarely used."

Key Steps & Outputs

The skill follows a clear execution flow:

1. Input Ingestion & Preprocessing: Accepts article pastes or file paths, performs deduplication and cleaning, and checks material quantity (minimum: ≥3 articles, ≥2000 words).
2. Dual-Engine Analysis & Quality Checkpoint (Phase 1 & 1.5): After deep analysis, it pauses to present an analysis quality summary for user confirmation.
3. Profile Synthesis & Verification (Phase 2 & 2.5): Synthesizes the thinking profile and writing style profile, cross-validates them to distill Thought DNA and Writing DNA, and generates a Prohibition List.
4. Output Generation (Phase 3): Generates two core artifacts:
   • A: Thought + Writing Style Prompt: A standalone prompt that incorporates thinking framework constraints.
   • B: Thought + Writing Persona Skill: A structured skill file (SKILL.md) with a built-in Dual Re-inspection System:
   • Re-inspection 1 (Writing Fingerprint): Checks if the output adheres to the fingerprint in vocabulary, sentence structure, rhetoric, sentiment, perspective, tone, etc.
   • Re-inspection 2 (Thought Consistency): Checks if the output aligns with the author's stance, emotion, argumentation style, identity, and taboos. Only if both checks pass is the output finally generated.
5. Blind Testing & Refinement (Phase 4 & 5): Validates by generating test texts and comparing them against the original, performing multi-dimensional matching assessments. After passing, it undergoes further refinement by dual agents (user-perspective and skill-creator-perspective) to enhance operability and completeness.

Applicability & Caveats

• Dependent on High-Quality Text: Effectiveness directly hinges on the quantity and quality of input corpus. Insufficient material may lead to incomplete or inaccurate style profiles.
• Based on Text Analysis, Not Web Search: All analysis is based solely on user-provided text, with no internet searches. This ensures purity of output but means it cannot fill gaps for concepts not explicitly stated by the author in the texts.
• Cloning, Not Creation: The skill aims to replicate existing writing styles and thinking patterns. For entirely new topics the author never explored, the clone's performance depends on the generalizability of its "thought DNA," which may introduce deviations.
• Complexity & Iteration: The full distillation process involves extensive analysis and verification, often requiring user confirmation and adjustments at key checkpoints to ensure the final output meets expectations.