Skip to content

AGENTS.md

Latest commit

 

History

History
45 lines (33 loc) · 3.14 KB

AGENTS.md

File metadata and controls

45 lines (33 loc) · 3.14 KB
Raw

AGENTS.md

Project scope

This repository maintains a curated catalog of SDN/NFV network resource allocation papers. Keep changes focused and preserve curated metadata unless authoritative evidence supports a correction.

Sources and generated files

  • data/papers.csv is the canonical paper catalog.
  • data/taxonomy.csv separates paper type, method families, and scenarios while categories remains the compatibility field used by the README.
  • data/venues.csv is the canonical venue/rank registry; materialized ranks values in papers.csv must match it.
  • data/provenance.csv records source and verification status for every paper. Use legacy unless a source and review date are known.
  • data/screening.csv records Include, Pending, and Exclude decisions so rejected candidates are not repeatedly reviewed.
  • data/README.header.md and data/README.footer.md are manually maintained README fragments.
  • data/README.md documents the schemas, allowed values, and ownership boundaries for the catalog sources.
  • docs/repository-settings.md records the required branch-protection, Pages, and scheduled-maintenance settings that cannot be enforced by repository files alone.
  • README.md, assets/papers.js, papers.html, robots.txt, sitemap.xml, visualization/analysis.md, and the SVG charts are generated artifacts; do not edit them directly.

Paper data rules

  • Preserve the existing CSV schema and quote every field.
  • Separate multiple values with ; and put the primary category or direction first.
  • Classify surveys with paper_type=survey_analysis; do not use the Survey section for primary research. Classify research methods independently in method_families.
  • Use the latest official CCF directory for CCF ranks and re-audit venue mappings when a new edition is released.
  • When both CCF and JCR ranks apply, put CCF first (for example, CCF-C; JCR-Q1). Treat JCR ranks as supplemental and do not add or update them without a source and ranking year.
  • Use unique doi:<doi> or arxiv:<id> paper IDs with matching canonical URLs.
  • Verify titles, authors, venue, year, identifiers, code, and data links from authoritative sources. Do not guess missing metadata; report uncertain items for user review.
  • Do not bulk reorder or reclassify papers without explicit approval.

Required refresh and validation

After changing any catalog CSV, README fragment, or generator, run:

python3 -m unittest discover -s tests -v
python3 scripts/validate_catalog.py
python3 scripts/generate_readme.py
python3 scripts/generate_readme.py --check

Commit source changes and regenerated artifacts together. A data update is incomplete if the freshness check fails. Before finishing, verify that the CSV record count matches the rendered README paper count.

The scheduled link workflow runs python3 scripts/check_links.py. A 404/410 is a catalog defect; timeouts, rate limits, and access-denied responses require manual review before changing a URL.

Visualizations must derive only from the validated catalog sources, remain deterministic and readable in light and dark themes, and state that they describe catalog coverage rather than total field output.