| name | dgm-md-to-pdf-chapter-polisher |
| description | Use when ChatGPT/Codex needs to transform multiple Markdown working files for a single research paper chapter into a polished, journal-style chapter draft, create export-ready Markdown/HTML variants, and generate a single-chapter PDF. |
DGM Markdown-to-PDF Chapter Polisher
Use this skill when the task is to turn several Markdown working files for one paper chapter into a coherent, publication-oriented chapter draft and export it as a single-chapter PDF.
This is a chapter-level skill, not a full-paper skill.
It is designed for Markdown-first academic workflows where a chapter starts as modular research files such as:
01_xxx.md02_xxx.md03_xxx.md- optional figures
- optional figure notes or captions
- optional reference or audit notes
When to use
Trigger this skill when the user asks to:
- merge several Markdown working files into one chapter
- polish a chapter into journal-style academic prose
- turn modular notes into a paper-facing chapter draft
- create a chapter export version
- generate a single-chapter PDF
- tighten overlap and repetition across several chapter source files
Typical chapter types:
- review / related work
- method
- evaluation design
- results
- discussion
Typical outputs:
manuscript/sections/<chapter_name>.mdmanuscript/sections/<chapter_name>_export.mdmanuscript/sections/<chapter_name>_export.html when usefuloutput/pdfs/<chapter_name>.pdf
What this skill is especially for
This skill is not a generic summarizer.
It should behave like a careful academic chapter polisher that:
- preserves the chapter's intellectual structure
- removes obvious overlap and repetition
- rewrites note-like text into coherent academic prose
- keeps claims no stronger than the source material supports
- produces a PDF-ready chapter without requiring the full paper to be finished
Core rules
- Work chapter by chapter, not full-paper by default.
- Read all chapter source files before merging.
- Identify overlap before rewriting.
- Preserve unique content even when compressing repeated ideas.
- Keep the chapter moderately polished, not overcompressed.
- Maintain stable terminology across the merged chapter.
- Do not hallucinate citations.
- Do not invent missing results, experiments, or findings.
- Preserve subsection numbering if it already exists and is useful.
- Use figure placeholders or figure references conservatively and only where structurally justified.
- Keep a clear distinction between:
- working notes
- merged working draft
- polished export
Standard workflow
- Inspect the source set.
- Identify the chapter folder or the explicit file list.
- Separate primary source files from support files such as captions, figures, notes, and audits.
- Map the structure.
- Determine the intended section logic and subsection hierarchy.
- Note repeated claims, repeated transitions, and conflicting terminology.
- Merge into a working chapter draft.
- Combine the source files into a single coherent Markdown chapter.
- Keep the structure explicit.
- Remove obvious duplication.
- Rewrite into paper-style prose.
- Convert note language into academic prose.
- Improve transitions across subsections.
- Keep paragraph length and density balanced.
- Add figure handling.
- Insert figure placeholders or direct figure references where they improve chapter flow.
- Keep captions separate if the repo workflow prefers separate caption files.
- Create export-ready chapter files.
- Prepare a polished export Markdown file.
- Create HTML only if it improves PDF conversion reliability.
- Generate the single-chapter PDF.
- Write the final PDF to
output/pdfs/.
- Validate heading hierarchy, spacing, margins, and figure placement.
Preferred chapter assembly logic
When merging modular source files:
- start from the strongest existing outline
- preserve the chapter's intended order
- merge repeated motivation paragraphs
- compress repeated definitions
- avoid repeating the same claim in slightly different language
- keep one best version of each transition
- keep caveats and non-claims explicit
When the source material is uneven:
- prefer a structurally clean draft over maximal source retention
- keep unresolved points visible with
TODO: rather than smoothing over real gaps
Style target
Target style:
- journal-style academic writing
- cohesive and readable
- moderately polished
- not too compressed
- not too verbose
- stable terminology
- strong paragraph transitions
- no product language
- no casual working-note voice in the final export
Figure handling
Support both patterns:
- placeholder style:
[Figure X about here]
- direct Markdown image references when the chapter export already uses local figure paths
If figures already exist:
- place them where they carry structural load
- do not over-explain what the figure already shows
- keep captions concise and academically worded
PDF guidance
This skill should coordinate with $pdf when final rendering and layout matter.
PDF expectations:
- single chapter only
- clean heading hierarchy
- readable margins and line width
- consistent spacing
- figures and captions aligned with chapter flow
- suitable for sharing as a chapter draft
Output discipline
Preferred outputs:
- merged working draft in
manuscript/sections/
- polished export draft in
manuscript/sections/
- optional HTML export in
manuscript/sections/
- final PDF in
output/pdfs/
If the source set is not ready for polishing:
- still produce a merged working draft
- state what blocked a stronger export
- do not fake completeness
Repository-specific guidance
This skill is designed for the decision-grade-memory workflow:
- modular chapter inputs first
- chapter-level assembly and polish second
- PDF export third
- full paper assembly later
Use it when chapter cohesion and export readiness matter more than raw ideation.
Supporting files
Use the companion templates in this skill when useful:
templates/chapter_workflow.mdtemplates/export_checklist.mdtemplates/chapter_output_map.md
