Writing Style Replicator

In content creation and analysis, accurately replicating a specific writing style is a complex challenge. It requires extracting stable linguistic patterns from existing text and converting them into actionable rules, rather than simple imitation or sentence-by-sentence copying. The core issue lies in distinguishing between accidental phrasing and stable style, ensuring the replication is verifiable and transferable, while avoiding inappropriate inferences about the author's persona.

The Specific Problem

Style replication faces multiple difficulties:
- Accuracy of pattern extraction: Style is a multifaceted concept encompassing vocabulary, syntax, structure, and thinking, not a single feature.
- Verifiability: Generated text must align with source text features, not just rely on subjective notions of "similarity."
- Adaptability and updates: Style models need to adapt to different tasks (e.g., stories, reviews, tutorials) and support incremental updates with new text, avoiding one-time fixation.
- Ethical boundaries: It must clearly avoid impersonating real authors, inferring sensitive attributes, and rely solely on observable textual evidence.

Core Workflow

The skill implements style replication through a systematic process, centered on an evidence-first analytical framework. Key steps include:
- Material ingestion and preprocessing: Supports multi-format text input, automatically identifies article boundaries, and performs cleaning and noise removal. Uses a four-dimensional scoring (volume, diversity, consistency, attributability) to assess sample quality, avoiding rigid thresholds.
- Text structuring and ten-dimensional analysis: Segregates text into L1 author commonality, L2 contextual characteristics, and L3 incidental characteristics. Then conducts ten-dimensional style analysis, each following the path: observation → evidence → frequency → explanation → rule → confidence. This covers dimensions like lexical fingerprint, syntactic skeleton, paragraph structure, rhetoric/imagery, emotion/attitude, narrative perspective, rhythm/sound, reasoning path, value orientation, and task adaptation.
- Evidence system and cross-validation: Each conclusion is bound to an evidence level (E3 high confidence, E2 conditional, E1 weak signal, U not included). Cross-mapping is established between thinking paths → language expression, language expression → text effects, and genre differences → conditional rules to build stable rules.
- Style model synthesis and output: Distills 5-8 observable core style genes, formulates avoidance rules, and generates an independent, decoupled Skill. This Skill includes core rules, genre modes, verification protocols, and can execute internal protocols like pre-writing reflection, generation, and dual-layer review.
- Verification and incremental updates: Uses five-item scoring (lexical fingerprint, syntactic rhythm, paragraph structure, tone/attitude, reasoning path) for structural evaluation, not pseudo-precise percentages. When new text is provided, incremental updates are performed with version change records, rather than overhauling long-term stable characteristics.

Boundaries and Caveats

This skill is not omnipotent; its applicability boundaries must be noted:
- Sample dependency: Insufficient samples (e.g., 1-2 pieces) yield only "local style snapshots," with reduced rule intensity. Multiple authors or genre mixtures require separate modeling or mode differentiation.
- No content copying: Strictly prohibited from misinterpreting unique source paragraphs, facts, or experiences as style features for replication. Style rules focus on the expression system, not the content itself.
- Evidence-level constraints: Single-inference (C-level) conclusions must not be written into mandatory rules to avoid over-generalization.
- Handling uncertainty: When contradictory features arise (e.g., style shifts over time), output applicable conditions rather than force-averaging, maintaining model transparency.
- Functional equivalence priority: When users request high-fidelity reproduction of a specific source text, prioritize generating a "functionally equivalent but phrasing-independent" version to ensure originality.
In summary, this is an evidence-driven, structured style analysis and replication tool, aimed at providing actionable, verifiable transformation of writing capabilities, not creating perfect clones.