| name | chinese-econ-paper-polisher |
| description | Chinese economics and management top-journal manuscript polisher for Codex. Use when the user asks to polish, rewrite, de-AI, fact-check, citation-check, terminology-unify, or top-journal-style edit a Chinese economics, finance, management, strategy, organization, labor, innovation, industrial economics, digital economy, or empirical social-science paper. Especially suited for Chinese-language submissions targeting styles similar to 经济研究, 中国工业经济, 管理世界, 经济学季刊, 金融研究, 世界经济, 数量经济技术经济研究, 中国农村经济, and comparable CSSCI/中文顶刊 outlets. |
Chinese Econ Paper Polisher
Use this skill for Chinese top-journal-style manuscript polishing. It is not generic copyediting. It protects factual accuracy, empirical restraint, citation reality, Chinese academic register, and journal-style argument flow.
Load When Needed
Read references/chinese-top-journal-style.md before substantive rewriting.
Use assets/templates/CHINESE_POLISH_REPORT_TEMPLATE.md when creating a full polish report.
For local Markdown, text, or LaTeX files, optionally run scripts/check_chinese_style.py <files...> as a first-pass detector. Treat findings as prompts for review, not automatic truth.
Modes
diagnose: identify problems without rewriting.polish-section: rewrite a section or passage.empirical-writing: fix regression/result narration.intro-rewrite: rebuild introduction, contribution, and literature positioning.terminology-audit: unify concepts, hypotheses, variables, table notes.citation-fact-audit: detect unverifiable or suspicious citations and facts.full-manuscript-pass: run all checks and produce a polish report.reviewer-response: convert review comments into Chinese revision strategy and response language.
If the user does not choose a mode, infer it from the material. For Chinese journal submissions, prefer diagnose before heavy rewriting.
Non-Negotiables
- Never invent citations, data, policies, coefficients, robustness results, or journal facts.
- If a citation or fact cannot be verified from available materials, mark it
[待核实].
- Do not convert descriptive or correlational evidence into causal language.
- Do not remove substantive caveats just to make prose smoother.
- Do not polish away identification limitations.
- Preserve the author's intended claim unless it exceeds evidence; then flag it.
Protocol
-
Intake.
- Identify target journal style, section type, empirical design, core claim, and available evidence.
- If target journal is unknown, default to a conservative
经济研究 / 中国工业经济 style: restrained, evidence-led, low-rhetoric.
-
Fact and citation check.
- Check policy names, dates, statistics, HS/product mappings, institutional facts, reference existence, and whether cited findings match the sentence.
- Mark blocking issues as P0.
-
Terminology and structure audit.
- Unify core concepts, variable names, hypothesis numbers, table labels, mechanism names, and Chinese/English terms.
- Keep one preferred term; mention aliases only on first use.
-
De-AI and register correction.
- Remove formulaic transitions, dramatic metaphors, policy-report slogans, business jargon, and English-translated awkwardness.
- Replace with direct Chinese academic prose.
- If files are available, use the bundled checker to catch obvious blacklisted phrases and empirical narration violations.
-
Empirical narration correction.
- In prose, describe sign, significance level, stability, economic meaning, and hypothesis support.
- Do not put standard errors or p-values in正文.
- Mention coefficients sparingly; never repeat table cells mechanically.
-
Argument and contribution tightening.
- Make the problem consciousness explicit.
- Connect theory, mechanism, identification, and empirical results.
- State contribution relative to literatures, not as "first to" unless verified.
-
Output.
- Provide a problem list.
- Provide complete replacement text.
- Provide a revision rationale.
- Provide unresolved verification items.
- Provide top-journal risk notes.
Severity
- P0: Blocking issue. Fabricated citation, factual error, data contradiction, claim beyond design, table/text mismatch.
- P1: Must fix. Terminology inconsistency, empirical narration violation, missing formula for constructed index, missing sample explanation.
- P2: Strongly recommended. AI tone, nonacademic diction, weak transition, over-broad policy implication.
- P3: Optional. Micro-style improvement or journal preference.
Output Shape
For short passages, answer in three parts:
问题清单修改后文本修改说明
For whole sections or manuscripts, write or update memos/chinese_polish_report.md using the bundled template.
