SKILL.md
Paper Style Extraction Skill
Goal: Reverse engineer academic style from PDF/TeX files and build the psmfiles/ knowledge base.
Workflow
- Locate Source Files:
Look for PDF, TeX, or TXT files in the current directory or user-specified path. If user requests new papers/datasets, use WebSearch to find them first.
- Extract Text:
Execute the extraction script: python paper-extract-style/extracttext.py <filepath> This creates refarticle/cleanedbody.md and refarticle/_appendix.md.
- Build Knowledge Base (Incremental Fusion):
Read: Process refarticle/cleanedbody.md. Update psmfiles/lexicondomain.md: Extract high-frequency sentence patterns for each section (Abstract, Intro, Methods, Experiments, Conclusion). Generalize: Replace specific nouns with placeholders (e.g., [Method], [Metric]). De-duplicate: Do not copy the static skeleton from LEXICON.md. Update psmfiles/DOMAINSKnowledge.md: Extract "Territory" statements (Industrial impact). Extract "Niche" statements (Critiques of prior work, Solved problems). Extract "Trends" and "Citations". Fusion Logic: Append new insights; merge duplicate views to reinforce arguments. Do NOT overwrite existing data. Generate psmfiles/STYLEGUIDE.md: Read paper-extract-style/TEMPLATES.md. Fill in the template based on observed style (Voice, Narrative Flow, Formatting). Statistics: Calculate average sentence length for each section. Appendix Check: Check *_appendix.md to define Appendix formatting standards.
- Completion:
* Report the location of generated assets in psmfiles/.