菜单
Build, query, and visualize literature relationship clusters using Kuzu and Cypher.
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
Fetch paper metadata from lightweight academic sources like arXiv and save normalized paper JSON records.
Write lightweight graph edges from normalized paper entities into a local graph store.
Build a lightweight local vector-style index record from normalized paper data.
Normalize extracted paper entities into canonical forms with lightweight audit logs.
Generate a lightweight structured reflection JSON for a single paper record.
Orchestrate the lightweight academic pipeline by chaining paper-fetch, paper-reflect, and paper-normalize.
| name | paper-cluster |
| description | Build, query, and visualize literature relationship clusters using Kuzu and Cypher. |
| requires | {"bins":["/root/openclaw-dev/envs/academic-embed/bin/python"]} |
Use this skill when the user wants to build or inspect a literature graph, run Cypher over paper and entity relationships, or render a paper cluster visualization.
/root/openclaw-dev/data//root/openclaw-dev/data/kuzu/academic_graph/root/openclaw-dev/data/visualizations//root/openclaw-dev/data/queries/last_query.jsonIngest one normalized paper:
/root/openclaw-dev/envs/academic-embed/bin/python /root/openclaw-dev/workspace/scripts/graph_engine.py ingest /root/openclaw-dev/workspace/data/normalized/2312.10793v3.json
Run a Cypher query directly:
/root/openclaw-dev/envs/academic-embed/bin/python /root/openclaw-dev/workspace/scripts/graph_engine.py query "MATCH (p:Paper)-[:MENTIONS]->(e:Entity) RETURN p.id AS paper_id, e.name AS entity LIMIT 10"
Render a paper cluster visualization:
/root/openclaw-dev/envs/academic-embed/bin/python /root/openclaw-dev/workspace/scripts/graph_engine.py visualize "MATCH (p:Paper)-[:MENTIONS]->(e:Entity) RETURN p.id AS source, 'Paper' AS source_label, 'MENTIONS' AS rel_type, e.name AS target, 'Entity' AS target_label LIMIT 50" --output-name paper-cluster-demo.png
query expects a Cypher string that has already been produced from user intent.visualize expects query output columns named: source, source_label, rel_type, target, target_label.