This framework addresses the challenge of adapting academic manuscripts to the specific writing conventions of target journals. It provides researchers and technical writers with a method to learn journal-specific styles from published papers and generate an auditable revision plan, enhancing manuscript alignment with publication venue expectations.

How It Works

The project employs a hybrid approach combining reusable "static" writing rules (e.g., discipline templates, anti-AI phrasing) with "dynamic" rules generated from a user-provided corpus. The dynamic layer analyzes papers from the target journal, supplemented by optional field-top or user-exemplar papers, to profile the journal's local writing culture. This profile informs a dynamic_writing_skill.md, which guides an agent through section-by-section manuscript revision, prioritizing hard constraints, then journal patterns, secondary sources, static skills, and finally cleanup rules.

Quick Start & Requirements

Installation for Claude Code involves copying the skill/* directory to ~/.claude/skills/journal-adapt/. For Codex, the skill/ folder can be integrated directly or referenced. The recommended workflow uses Markdown inputs for both the manuscript and corpus papers. A PDF workflow is supported but requires a reliable PDF-to-Markdown converter, with MinerU being an option that may present installation challenges. The primary invocation is via the ./journal-adapt command or through agent prompts. Output is saved in a [manuscript_name]_revised/ directory.

Highlighted Details

• Static Skills: Offers a range of pre-built static skills for disciplines like Economics, ML/CV/NLP, CS, and Philosophy, alongside general AI-writing cleanup rules. Custom static skills are also fully supported.
• Dynamic Corpus Prioritization: A clear priority system (P1-P5) governs rule application, ensuring hard constraints are paramount, followed by target journal patterns, secondary sources, static skills, and cleanup rules.
• Auditable Revision Framework: Generates an editable dynamic_writing_skill.md and style_profile.md, providing transparency and human oversight for the revision process, rather than fully automated rewriting.
• Corpus Sizing: Recommends 5-8 target journal papers, 2-5 secondary papers, and 1-3 user exemplars for robust dynamic skill generation.

Maintenance & Community

The README outlines potential contributions for improving installation paths, PDF conversion, and example corpora, but does not provide direct links to community channels, active contributor lists, or a public roadmap.

Licensing & Compatibility

The project is released under the MIT license, which is permissive and generally compatible with commercial use and closed-source projects.

Limitations & Caveats

This tool is strictly limited to English-language academic writing. The quality of PDF conversion is dependent on the chosen converter, and MinerU may encounter setup issues. The system extracts writing patterns and does not generate new facts, citations, or claims. The output dynamic skill requires human review before initiating manuscript revisions.