A conservative AI-writing risk detection and thesis rewriting skill for English and Chinese academic writing.

This Openclaw & Hermes-Agent skill helps identify and reduce AI-generated writing-risk patterns in master’s thesis text while preserving academic meaning, citations, variables, hypotheses, statistical results, and document structure.

It is designed for users who need careful, detector-informed academic revision, especially for thesis chapters affected by Turnitin AI, CNKI AIGC, or other external AI detection reports.

This is not a detector-bypass tool. It does not guarantee any AI detector score. It focuses on minimal, evidence-preserving revision and academic integrity.

AI writing detection, AIGC detector, thesis AI rewrite, academic AI detection, Turnitin AI, CNKI AIGC, AI-generated writing risk, reduce AI score, master thesis rewriting, English thesis editing, Chinese thesis AI risk, detector-informed rewriting, academic integrity, minimal-edit rewriting, AI-polished text, AI-refined text, thesis revision skill.

中文关键词：降AI率，降低AIGC检测率，知网AIGC，Turnitin AI检测，论文AI痕迹，硕士论文改写，英文论文降AI，中文论文AI风险，AI润色痕迹，学术论文改写。

The AIGC Detector & Rewriter Skill analyzes thesis text for AI-like writing patterns and revises selected high-risk passages through controlled, minimal edits.

It targets risks such as:

• Repeated sentence openings
• Over-smooth academic logic
• Mechanical paragraph structure
• Formulaic literature review patterns
• Repetitive hypothesis-development structure
• Uniform empirical-result reporting
• Generic conclusion and implication language
• AI-polished or AI-refined academic style
• Chinese template phrases and CNKI-sensitive writing patterns
• Qualitative research passages with decorative quotations or weak authorial interpretation

The skill supports:

• English thesis text
• Chinese thesis text
• Mixed English-Chinese thesis documents
• Quantitative research
• Qualitative research
• Mixed-methods research
• External AI detector report analysis
• Chapter-by-chapter revision workflow
• DOCX / Markdown / text-based workflows

This skill follows one central rule:

Reduce AI-writing risk by minimally editing the original thesis text, not by regenerating it.

The skill does not rewrite a chapter from scratch. It does not replace the author’s work with fluent AI-generated prose. Instead, it works sentence by sentence and paragraph by paragraph to reduce detectable writing patterns while keeping the original academic content intact.

Allowed operations include:

• Substituting individual words
• Reordering clauses where safe
• Splitting overly balanced sentences
• Recasting repeated sentence openings
• Reducing formulaic transitions
• Making existing reasoning more visible
• Reorganizing paragraph architecture without changing meaning
• Relocating existing author reasoning when traceable
• Preserving or increasing word count when required

Forbidden operations include:

• Whole-paragraph regeneration
• Whole-chapter rewriting
• Back-translation
• Detector-bypass rewriting
• Third-party humanizer rewriting
• Fabricating data, citations, examples, quotes, or researcher voice
• Changing variables, hypotheses, coefficients, p-values, or conclusions
• Claiming that an external detector score has improved without a comparable external re-test

Many AI rewriting tools make academic text more fluent, smoother, and more polished. That can make the text look more AI-generated, not less.

This skill is built around the opposite strategy:

• Keep as much original author wording as possible
• Avoid over-polishing
• Preserve non-native but acceptable academic style
• Break repeated structures without deleting content
• Treat detector scores as external evidence, not as something the model can predict
• Work chapter by chapter with external re-testing

The goal is not to create perfect native-speaker prose. The goal is to preserve the author’s thesis while reducing structural AI-writing risk.

The skill checks AI-risk signals at different levels:

English and Chinese thesis text are handled differently.

This prevents Chinese thesis text from being incorrectly judged by English-only AI-writing indicators.

The skill can use external detector reports as calibration signals, including:

• Turnitin AI reports
• CNKI AIGC reports
• GPTZero-style reports
• Originality-style reports
• Highlighted AI-risk fragments
• Chapter-level AI scores
• Whole-document AI scores

However, the skill does not claim to simulate or predict any detector.

Internal risk scores are only used for diagnosis and prioritization. Only a real external re-test can confirm whether the detector score changed.

The rewriting process protects:

• Original academic meaning
• Research logic
• Variables and abbreviations
• Hypotheses
• Citations
• Reference names
• Statistical values
• Coefficients
• p-values
• Table and figure numbers
• Methodological claims
• Qualitative quotes
• Participant meaning
• Chapter and paragraph structure

Every rewritten paragraph must pass quality gates before being accepted.

The recommended workflow is:

1. Build a chapter version ledger
2. Record current external detector scores
3. Select the highest-risk chapter
4. Diagnose paragraph and section risks
5. Rewrite only the selected high-risk areas
6. Run quality gates
7. Ask the user to re-test the chapter externally
8. Accept the new version only if the external re-test improves
9. Continue to the next target chapter only after confirmation

A chapter is considered done only when its measured external AI score is below the target threshold.

Default target: below 30%, unless the user specifies another threshold.

This skill is useful for:

• Master’s thesis AI-writing risk reduction
• English thesis revision for non-native authors
• Chinese thesis CNKI AIGC risk review
• Turnitin AI report response
• Chapter-level AI detector remediation
• Literature review de-templating
• Hypothesis-development rewriting
• Methodology section revision
• Empirical results reporting improvement
• Discussion and conclusion rewriting
• Qualitative interview findings improvement
• Mixed-methods thesis editing
• Detector-informed academic editing workflows

This skill is not:

• An AI detector
• A detector bypass tool
• A plagiarism reduction tool
• A similarity-score reduction tool
• A third-party humanizer
• A full thesis generator
• A citation generator
• A tool for inventing research evidence
• A tool for making fake “human-like” errors
• A guarantee of Turnitin, CNKI, or any detector result

It does not encourage academic misconduct. It is designed for conservative academic revision of user-provided thesis text.

For best results, provide one of the following:

Use this for quick diagnosis or light rewriting.

Recommended input:

Please review this paragraph for AI-writing risk and revise it minimally.
[Paste paragraph here]

Use this for chapter-level diagnosis and rewriting.

Recommended input:

Chapter: Chapter 4 Results
External detector score: 42%
Target score: below 30%
Detector: Turnitin AI
Please diagnose and revise only the high-risk paragraphs.

Use this when you have highlighted AI-risk sections.

Recommended input:

I have uploaded the Turnitin/CNKI AI report.
Please identify the highest-risk sections and create a chapter-by-chapter revision plan.
Do not rewrite the whole thesis at once.

Recommended input:

请按照中文论文AI痕迹风险规则检查这一段，不要套用英文D1–D17评分。
请最小改写，不要删减字数，不要改变原意。
[粘贴中文段落]

Depending on task size, the skill may produce:

• Scope label
• Hit risk dimensions
• Selected techniques
• Protected elements check
• Quality gate result
• Revised paragraph
• Major changes made

• Task routing result
• External detector evidence map
• Protected elements list
• Paragraph risk table
• Dimension-to-technique execution plan
• Three-pass rewrite evidence
• Revised version
• Phase 7 quality gate result
• Word-count preservation check
• External re-test instruction
• Next-round status

When auditing the Skill document itself, the output focuses on:

• Contradictions
• Missing rules
• Execution risks
• Ambiguous instructions
• Workflow conflicts
• Quality gate weaknesses
• Priority fix list

The skill follows these rules strictly:

1. Do not fabricate sources, data, findings, quotes, or researcher reflections.
2. Do not change protected academic elements.
3. Do not rewrite whole chapters as newly generated prose.
4. Do not use back-translation.
5. Do not route thesis text through third-party humanizer tools.
6. Do not claim external detector improvement without comparable external re-test evidence.
7. Do not over-polish the author’s style.
8. Do not intentionally add grammar or spelling mistakes.
9. Do not delete content to reduce AI-risk patterns.
10. Do not merge paragraphs or damage document structure.

Please revise the following thesis paragraph to reduce AI-generated writing patterns.
Keep the meaning, citations, variables, and academic logic unchanged.
Do not over-polish the language.
Do not shorten the paragraph.

Turnitin AI marked Chapter 5 as 48%.
Please create a chapter-level risk diagnosis first.
Then revise only the highest-risk paragraphs using minimal edits.
Do not claim the score is reduced until I re-test it.

请按照中文论文AIGC风险检查规则，检查下面内容是否存在模板化表达、空泛总结、机械连接和中文AI痕迹。
请只做最小改写，不要改变研究结论，不要删减字数。

Please review this Skill document for contradictions, missing rules, ambiguous instructions, workflow risks, and quality-gate weaknesses.
Do not rewrite the whole document.
Only identify issues and propose precise fixes.

1. Upload thesis chapter or detector report
2. Identify language route: English / Chinese / mixed
3. Identify research type: quantitative / qualitative / mixed-methods
4. Record external detector score if available
5. Build chapter version ledger
6. Diagnose paragraph-level and section-level AI-risk patterns
7. Select only the highest-risk paragraphs
8. Rewrite minimally
9. Run quality gates
10. Re-test externally
11. Accept or discard candidate version based on comparable re-test result

Each revised paragraph or chapter should pass these checks:

If any gate fails, the revision is incomplete.

Suggested repository structure:

aigc-detector-rewriter/
├── README.md
├── SKILL.md
└── references/
    ├── detection_principles.md
    ├── rewrite_methods.md
    ├── qualitative_authorship_restoration.md
    └── chinese_text_ai_risk.md

This repository is intended to be used as a custom AI writing skill or prompt-based academic revision module.

Basic usage:

1. Copy SKILL.md into your skill environment.
2. Keep all reference files in the references/ folder.
3. Provide thesis text, chapter text, or detector reports as input.
4. Run the task according to the routing rules.
5. Re-test rewritten chapters externally before accepting a new baseline.

AI detector results can vary across tools, dates, settings, and document formats.

This skill treats detector results as external evidence, not as a guaranteed truth.

The skill may help reduce writing patterns that detectors often flag, but it cannot promise that:

• Turnitin AI will report a specific score
• CNKI AIGC will fall below a specific threshold
• Any detector will classify the text as human-written
• A rewritten version will always score lower

A rewritten chapter is only a candidate version until a comparable external re-test confirms improvement.

Use this skill only for legitimate academic revision.

Appropriate uses:

• Improving clarity while preserving meaning
• Reducing formulaic writing
• Removing excessive AI-polished patterns
• Protecting thesis originality
• Making the author’s actual reasoning more visible
• Reviewing detector-highlighted areas conservatively

Inappropriate uses:

• Submitting fabricated research
• Hiding ghostwritten work
• Manufacturing fake human style
• Bypassing academic integrity systems
• Replacing the author’s thesis with AI-generated prose

No. It can reduce AI-writing risk patterns, but only an external re-test can confirm the actual detector result.

No. The recommended workflow is chapter by chapter. Full-thesis batch rewriting increases the risk of over-editing, structure damage, and detector regression.

Yes. Chinese thesis text uses a separate Chinese AI-risk route rather than the English D1–D17 framework.

Yes. It includes qualitative-specific checks for interviews, themes, quotations, participant meaning, researcher voice, and evidence-based interpretation.

Yes. External detector reports can be used for calibration and prioritization, but the skill does not claim to reproduce detector logic.

No. This skill explicitly rejects third-party humanizer logic. It focuses on academic integrity, minimal edits, and source-traceable revision.

No. It may preserve acceptable non-native academic style, but it never creates deliberate grammar or spelling errors.

Choose a license that fits your intended use.

Recommended options:

• MIT License for open use
• Apache-2.0 for open use with patent language
• CC BY-NC 4.0 if you want non-commercial reuse only

This project is for academic writing support and document quality improvement. It does not provide legal, institutional, or academic-integrity advice. Users are responsible for following their university’s rules, disclosure requirements, and research ethics standards.

External AI detector results are not guaranteed. Always verify revisions with the relevant institution-approved process.

A conservative AI-writing risk detection and thesis rewriting skill for English and Chinese academic writing, supporting Turnitin AI, CNKI AIGC, detector-informed revision, minimal-edit rewriting, and academic integrity protection.

aigc
ai-detection
turnitin-ai
cnki-aigc
academic-writing
thesis-writing
thesis-revision
ai-writing-risk
english-thesis
chinese-thesis
academic-integrity
prompt-engineering
custom-skill
writing-skill
detector-informed-rewriting