paper-material-package

Paper Material Package

Purpose

Create ref_papers/paper-notes.md as the readable handoff between Search and downstream specialists. Keep each paper separate and factual so Writing can build cross-paper taxonomy and synthesis without guessing what the sources say.

This Skill does not search for candidates, verify metadata by itself, download PDFs, or convert documents. Use paper-search, arxiv, and pdf-to-markdown first. It does not write a literature review, compare papers across the collection, choose a taxonomy, or draft manuscript prose.

Inputs

Inspect these exact-cwd artifacts when available:

• ref_papers/source.json
• ref_papers/pdf/
• ref_papers/text/
• an existing ref_papers/paper-notes.md
• the dispatch topic, inclusion scope, and expected paper count

Follow another existing layout only when the dispatch explicitly identifies it. Do not infer paper facts from search snippets when readable source text exists.

Stable Paper Keys

Give every selected paper one stable key, using the first available value:

1. arXiv id, including version only when version-specific evidence matters;
2. DOI;
3. OpenReview forum id;
4. another verified repository identifier;
5. a short normalized title key when no stable identifier exists.

Store the key as note_key on the matching source.json entry and use it in note headings, citation checks, and later Writing tasks. Preserve every existing manifest field. Do not invent an identifier.

Output Contract

Maintain one file at ref_papers/paper-notes.md:

# Paper Material Notes

## Material Scope
- Topic:
- Intended use: survey | empirical | hybrid | citation-only
- Requested paper count:
- Selected paper count:
- Full-text-ready count:
- Sources and date range:
- Known coverage gaps:

## <stable-key>: <verified title>
- Citation metadata: <authors; year; venue; DOI/arXiv/OpenReview URL>
- Local sources: <PDF path; readable text path>
- Material status: complete | partial
- Research question: <what problem the paper addresses>
- Method: <factual mechanism or study design>
- Data and evaluation: <datasets, tasks, baselines, metrics, protocol>
- Main findings: <results supported by the source>
- Limitations: <reported or directly evidenced limitations>
- Topic relevance: <why this paper belongs in the requested material set>
- Evidence locators:
  - <section/page/table/figure/text heading>: <fact supported there>
- Unresolved material gaps: <none or exact missing evidence>