eda-paper-fetch

name: eda-paper-fetch description: Fetch primary-source paper metadata when evidence is missing or weak for EDA/model/flow claims. Use for requests like building reproducible paper candidate lists, generating user-download queues, and recording citation metadata for local validation.

EDA Paper Fetch

Overview

Use this skill to produce evidence-backed paper sets, not ad-hoc links. Always output a local manifest and evidence note.

Workflow

1. Define scope and constraints.

• topic and claim to support/refute,
• date window and venue preference,
• target count (default: 5-15 papers).

2. Search primary sources first.

• Load references/source-priority.md.
• Prefer official sources (publisher/arXiv/OpenReview/DOI landing pages).
• Do not claim availability without opening source pages.

3. Triage and score.

• Keep only papers relevant to the exact claim.
• Record reason for inclusion/exclusion using references/triage-template.md.

4. Generate user-download queue instead of auto-download.

• Write docs/papers/queues/<topic_slug>.download_queue.tsv.
• Include columns: paper_id, title, url, doi_or_arxiv, preferred_source, access_type(open/paywalled).
• User manually downloads PDFs to docs/papers/pdf/ using filename:
  • <year>_<first_author>_<short_title>.pdf.

5. Produce manifest and summary.

• Write docs/papers/manifests/<topic_slug>.paper_manifest.tsv.
• Write docs/papers/manifests/<topic_slug>.paper_manifest.md.
• Include fields: paper_id, title, authors, venue, year, url, doi_or_arxiv, local_pdf_path, relevance_note, status(todo/downloaded/summarized).

Output Contract

Always return:

1. local manifest path,
2. local download queue path,
3. unresolved paywalled items,
4. one-paragraph evidence fitness assessment.

References

Read only when needed:

1. references/source-priority.md
2. references/triage-template.md