菜单
Search Semantic Scholar for papers with rich metadata — citation counts, AI-generated TLDRs, influence scores, citation graphs. Use for finding established research, tracing citation chains, understanding paper impact and field structure.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Web-grounded SOTA discovery using Perplexity Sonar models via OpenRouter. Use when academic APIs return foundational papers but you need to know what's CURRENT — what has superseded older methods, what practitioners actually use today, and what was released in the last weeks/months.
AI frontier plugin — gives Claude and agents access to latest scientific research, SOTA discovery, and architecture evaluation. Use when needing SOTA methods, research consensus, implementation guidance, or evaluating whether a codebase uses optimal patterns and architectures.
Search arXiv for latest AI/ML preprints. Use when Claude needs bleeding-edge research not yet in peer-reviewed journals — newest methods, architectures, training techniques, benchmarks.
Search Hugging Face Papers for AI/ML research with AI-generated summaries, keywords, community upvotes, and trending papers. Use when Claude needs community-curated research, AI-generated paper summaries, or trending ML papers.
| name | semantic-scholar-search |
| description | Search Semantic Scholar for papers with rich metadata — citation counts, AI-generated TLDRs, influence scores, citation graphs. Use for finding established research, tracing citation chains, understanding paper impact and field structure. |
| keywords | ["semantic-scholar","citations","papers","research","tldr","impact","academic","literature"] |
Search 200M+ papers with rich metadata. No API key required (100 req/sec).
Two modes:
SCRIPT=$(find ~/.claude/plugins -path "*/ai-frontier/skills/semantic-scholar-search/scripts/search.mjs" 2>/dev/null | head -1)
# Search mode
node "$SCRIPT" "your query" [maxResults] [--year=2024] [--open-access]
# Detail mode (get paper + citations + references)
node "$SCRIPT" --detail=<s2PaperId|DOI|ArXiv:id>
| Arg | Default | Description |
|---|---|---|
| query | (required) | Natural language search query |
| maxResults | 10 | Number of results (1-100) |
| --year=YYYY | none | Only papers from YYYY onward |
| --open-access | false | Only papers with free PDF |
| --detail=ID | n/a | Get full details for a specific paper |
results[].tldr — AI-generated one-sentence summary (from S2)results[].citations — total citation countresults[].source_specific.influential_citations — citations from influential papersresults[].source_specific.fields_of_study — e.g., ["Computer Science", "Mathematics"]paper.top_references and paper.recent_citations for citation graph# Find influential papers on RLHF
node "$SCRIPT" "reinforcement learning from human feedback" 10
# Recent papers only (2025+)
node "$SCRIPT" "mixture of experts scaling" 15 --year=2025
# Get details + citation graph for a specific paper
node "$SCRIPT" --detail=ArXiv:2210.11416