Academic Paper Formatter

Use this skill for Chinese domestic undergraduate, master's, and doctoral thesis .docx files. The school-provided template is the formatting authority: analyse it first, generate a rules JSON, then format a separate working copy of the thesis. Do not replace this flow with an assumed university standard.

Choose a degree level only to surface a manual structural checklist. It never adds a required section or overwrites a template-derived format:

python scripts/analyze_docx.py school-template.docx --degree master --output school-rules.json

Required workflow

1. Keep the original template and thesis untouched. Work on copies only.

2. Generate rules from the school's template:

   python scripts/analyze_docx.py school-template.docx --output school-rules.json
   

3. Inspect school-rules.json. Confirm ambiguous roles such as cover title, Chinese/English abstracts and keywords, declaration page, contents, heading levels, captions, references, acknowledgements, appendices, and degree-period research achievements against the template. These components vary by school and degree; edit only confirmed values and record unresolved warnings.

4. Validate the generated rules:

   python scripts/validate_format.py school-rules.json
   

5. Format a copied thesis and then validate it:

   python scripts/format_document.py thesis-source.docx thesis-formatted.docx --rules school-rules.json
   

python scripts/validate_format.py school-rules.json --document thesis-formatted.docx

For a review-only domestic degree structure check, add `--degree undergraduate`,
`--degree master`, or `--degree doctoral`. Use `--strict-degree` only after the
school's written requirements confirm that every flagged component is mandatory.

6. Render the resulting DOCX and visually inspect every page before delivery.
   Check page geometry, Chinese font rendering, page breaks, table clipping,
   captions, headers/footers, and reference wrapping.

## Advanced workflow

### Content completion

First locate placeholders. Draft proposed text in a separate review file or in
conversation; do not silently generate or insert academic claims. After the
user approves the exact mapping, apply it to a new DOCX copy:

```bash
python scripts/complete_content.py thesis.docx --report placeholders.json
python scripts/complete_content.py thesis.docx --replacements approved-content.json --output thesis-completed.docx

approved-content.json is an object mapping the exact placeholder text to the approved replacement. The tool preserves paragraph structure and reports placeholders split across Word runs instead of replacing them unsafely.

Citations and references

Audit numeric citations against the existing bibliography before changing it:

python scripts/manage_references.py thesis.docx --rules school-rules.json --report references-audit.json

The report finds missing entries, uncited entries, and duplicate labels. It never fabricates bibliographic metadata. Supply vetted records as --metadata-json verified-references.json --output ... to insert missing entries automatically. --renumber --output ... is allowed only after the audit is clean and only renumbers existing numeric citations and entries.

Floating figures and Word fields

Normalize existing floating-image anchors without modifying the image bytes or captions, then request Word to update field results when the output opens:

python scripts/normalize_floating_figures.py thesis.docx thesis-figures.docx --horizontal center --vertical top
python scripts/update_fields.py thesis-figures.docx thesis-final.docx

update_fields.py updates the document setting used by Word for TOC, REF, and PAGEREF fields. Open the result in Word and accept its update prompt; render it again afterwards because page-number and TOC changes affect layout.

PDF formatting guides

When the school supplies a PDF guide in addition to a DOCX template, extract page-cited evidence first:

python scripts/parse_pdf_rules.py school-guide.pdf --output pdf-requirements.json

Review the evidence, inspect diagram-only or scanned pages visually, and merge only confirmed requirements into school-rules.json. A PDF extractor must not infer layout rules from unverified text.

Focused tools

• scripts/format_heading.py: apply only title and heading role rules.
• scripts/format_tables.py: apply table font, alignment, and cell alignment.
• scripts/format_figures.py: apply figure/table caption rules.
• scripts/format_references.py: format the reference heading and entries.
• scripts/parse_rules.py: inspect generated rules or regenerate them from a template.
• scripts/complete_content.py: find placeholders and apply approved content.
• scripts/manage_references.py: audit or safely renumber numeric citations.
• scripts/normalize_floating_figures.py: make anchored drawings non-overlapping.
• scripts/update_fields.py: mark TOC/cross-reference fields for updating on open.
• scripts/parse_pdf_rules.py: extract reviewable, page-cited rules from a PDF guide.

Read references/docx-formatting.md before formatting. For citation-list scope, also read references/citation-rules.md. For degree-level structure, read references/chinese-degree-thesis-workflow.md.

Safety and fidelity

• Do not alter the original template or source paper.
• Do not invent school requirements, citations, captions, reference entries, or content. Report anything the template does not reveal.
• Preserve text, field codes, section breaks, tracked changes, images, and table content. This skill changes layout and styles only.
• The JSON files in templates/ are generic starting points only. Never present them as a school's submission standard when an official template is available.