Expand References

Expand References

Turn one to three seed papers into a structured follow-up reading list. Use this when the human already has anchor papers and wants the next papers to read.

Arguments

• Positional arguments are the seed papers. Quote multi-word titles.
• --negative may be repeated to push the workflow away from an unwanted cluster.
• --pool all-cs|recent selects the Semantic Scholar recommendation pool.
• --limit controls how many raw recommendations are requested before reranking.
• --per-bucket-limit caps each curated bucket after scoring.

Workflow

1. Run python scripts/run.py ....
2. Read result.closest_neighbors for the immediate next reads.
3. Read result.bridge_papers for papers that connect multiple seeds.
4. Read result.foundational, result.methodological, result.recent, and result.surveys_or_benchmarks for curated slices of the neighborhood.
5. If the result is sparse or off-topic, adjust the seed set or add --negative papers and rerun.

Output

• The script prints the unified JSON envelope described in output_contract.md.
• The underlying workflow result is ExpandReferencesResult.to_dict().
• result.notes captures dropped records and other execution notes.

When To Escalate

• Fewer than one clear seed paper is available.
• The resolved seeds are obviously duplicates or wrong papers.
• The output is empty even after trying better seeds or a different recommendation pool.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: zongmin-yu
• Source: zongmin-yu/semantic-scholar-skills
• License: MIT

Install and usage instructions live in the source repository linked above.