SKILL.md
ccf-rank
Use this skill to answer CCF ranking questions quickly and consistently.
Data source
- Dataset index:
references/manifest.json
- Year data:
references/<year>/rankings.json
- Year exclusions:
references/<year>/excluded_venues.json
Query workflow
- Run the lookup script with Node.js (preferred for installable skill compatibility):
``bash node scripts/queryccfrank.mjs "<venue name>" ` It uses the latest available year by default. Add --year <YYYY>` to query a specific year.
- For ambiguous venue names, narrow by type/rank:
``bash node scripts/queryccfrank.mjs "<query>" --year 2026 --type conference --rank A --top 20 ``
- If user asks for several venues, run the script per venue and return a compact table with:
venue, type, rank, area, url.
- If there are multiple high-score matches, show the top matches and explicitly ask user to disambiguate.
Output rules
- Always include
type (conference or journal) and rank (A/B/C).
- Include the CCF
area category when available.
- Include the canonical venue name and DBLP URL from the dataset.
- If no confident match is found, say so explicitly and list closest candidates.
Update workflow (new CCF version)
- Replace the PDF file with the newer official CCF version.
- Rebuild one year dataset:
``bash python scripts/buildccfdataset.py "/absolute/path/to/new.pdf" --year 2028 ``
- Update
references/<year>/excluded_venues.json if the PDF has deleted entries that text extraction cannot reliably detect from styling.
- Spot-check representative venues (for example
ICML, NeurIPS, CVPR, TOCS) using queryccfrank.mjs --year <year>.