qinyan-nature-writing

Qinyan Nature Paper Writing

Treat the paper as an auditable chain of argument, not a set of sentences that "look like a top journal." Lock down the facts and evidence first, then arrange the reader's path, and write last.

Scope of Work

• Handle drafting, restructuring, and first-submission materials from materials to paper sections.
• Leave pure language polishing to qinyan-nature-polishing。
• Leave pre-submission peer review to qinyan-nature-review。
• Leave statistical design and reporting verification to qinyan-nature-statistics。
• Leave scientific figure design and export to qinyan-nature-figures。
• Do not generate results, mechanisms, statistical significance, references, experimental details, or journal policies that do not exist.

Writing Foundation That Must Be Established First

Extract from the user's materials and clearly mark:

1. Central claim: The core proposition that the paper actually proves or supports.
2. Evidence set: The figures, tables, experiments, models, datasets, and statistical results that support the proposition.
3. Boundary: The extrapolation scope, mechanistic explanations, or causal conclusions that the evidence cannot support.
4. Audience: Why cross-disciplinary readers should care, and what specialist readers need to see.
5. Terminology ledger: The single standard wording for methods, datasets, metrics, abbreviations, and symbols.
6. Unresolved facts: Facts that still require author confirmation; uniformly write as AUTHOR_INPUT_NEEDED。

If the core proposition, key evidence, or boundaries are missing, first provide a framework that can continue the work and a gap list, without fabricating body facts.

Execution Process

1. Classify the task. Determine the paper type, target section, target journal, source language, word limit, and delivery format.
2. Build the argument spine. Express "research object—key advance—method—evidence—applicable boundaries" in one sentence.
3. Establish an evidence matrix. Bind each major claim to at least one evidence location; downgrade, delete, or mark as pending any claim without evidence.
4. Design section responsibilities. Let each paragraph carry only one main function: background, gap, method, finding, comparison, explanation, significance, or limitation.
5. Draft in evidence order. Write the results and figure logic first, then the methods and discussion, and finally converge on the abstract and title; if the user specifies another order, follow the user.
6. Calibrate strength of expression. Distinguish observational, correlational, predictive, interventional, and mechanistic evidence; make verb strength consistent with the evidence level.
7. Run a structural audit. For text files, execute python scripts/manuscript_audit.py <file>, and handle blockers and warnings.
8. Deliver and expose gaps. Provide paste-ready body text, key editing notes, evidence risks, and items for the author to confirm.

When section-level structure is needed, read references/section-blueprints.md. When a complete argument and evidence mapping method is needed, read references/argument-workflow.md. When first-submission materials are needed, read references/submission-package.md。

Default Output

写作设定
- 论文类型 / 章节 / 目标期刊：
- 核心命题：
- 证据边界：

论证与段落地图
- P1:
- P2:

可粘贴正文
[draft]

证据与措辞风险
- [claim] → [evidence pointer / missing]

AUTHOR_INPUT_NEEDED
- [仅列事实性问题]

When the user requests only body text and the materials are sufficient, simplify the output, but still retain necessary risk markers.

Quality Threshold

• Every major claim can be traced back to evidence provided by the user.
• The results, directions, and numerical values in the abstract are consistent with the main text.
• The gap raised in the introduction is genuinely addressed by this study.
• The results section does not pass off interpretation as observation, and the discussion section does not introduce new results that were not presented.
• The limitations specifically state the applicable boundaries and are not written as a polite closing sentence.
• The title and abstract avoid unverifiable expressions such as "first," "breakthrough," and "universal."
• The author, affiliation, conflict of interest, and recommended reviewer information in the initial submission materials are all confirmed by the authors.

Material routing

Task	Read
Build a claim–evidence map, argument spine, or terminology ledger	references/argument-workflow.md
Draft or restructure specific sections	references/section-blueprints.md
Cover letter, title page, declarations, and submission checks	references/submission-package.md