Teach an LLM to write like your target journal

What it does

journal-adapt-writing-skill is a Claude/Codex skill that ingests published papers from your target journal, distills their rhetorical habits into an editable dynamic_writing_skill.md, and uses that profile to rewrite your manuscript section by section. It treats journal style as a learnable corpus signal rather than a vague instruction.

The interesting bit

The framework layers a strict priority stack: hard facts and citations are immutable, target-journal patterns take precedence, and static discipline rules only apply when the corpus is silent. This keeps the LLM from hallucinating citations while still letting it mimic how your destination journal frames contributions or structures introductions.

Key highlights

• Generates an auditable, human-editable style profile before touching your manuscript
• Accepts Markdown or PDF input, though PDF conversion via MinerU can be brittle
• Supports optional static base skills (economics, ML, CS, philosophy) plus user or lab exemplars
• Outputs section-by-section revision logs so you can see why the agent changed a paragraph
• Explicitly refuses to add facts, citations, or results not already present in your draft

Caveats

• English-language academic writing only; no support for other languages
• Requires human review of the generated dynamic skill before revision begins
• PDF-to-Markdown conversion depends on external tools and may fail on some local setups

Verdict

Worth a look if you are resubmitting the same manuscript to multiple journals and tired of manually retuning tone. Skip it if you want the LLM to generate original research or citations for you.