| name | citation-graph-explorer |
| description | Explore one-hop references and related works from a DOI or OpenAlex work, then produce a bounded, provenance-aware citation neighborhood for literature expansion or baseline discovery. Use when a researcher wants to snowball from seed papers without treating citation count or graph proximity as scholarly quality. |
| tools | ["read_file","write_file","process_seed_paper","fetch_outgoing_citations","build_domain_map","finish_task"] |
| strict_tools | true |
| model_tier | standard |
| temperature | 0.1 |
| allowed_read_prefixes | ["user_inputs/citation-graph-explorer/","user_seeds/","literature/","_runtime/skill_sessions/"] |
| allowed_write_prefixes | ["user_seeds/","literature/"] |
| outputs_expected | {"graph_report":"literature/skill_citation_graph.md","graph_record":"literature/skill_citation_graph.json","domain_map":"literature/skill_citation_domain_map.json"} |
| interaction | {"mode":"guided","language":"zh-CN","summary":"从 DOI 或 OpenAlex 论文标识出发抓取一跳参考文献和 related works，输出可追溯的扩展候选与图谱边界，不把图谱位置当成质量结论。","request_required":true,"request_prompt":"请说明要探索的主题、希望发现的角色（基础方法、基线、数据集、桥接理论等）和每个种子允许的最大邻居数。","example_request":"从这些 DOI 扩展当前研究主题的基础方法和竞争性 baseline；每个种子最多 30 条参考文献，单独标出只有元数据的候选。","required_inputs":[{"id":"seed_identifiers","label":"DOI 或 OpenAlex 种子清单","description":"一行一个 DOI、doi.org URL、OpenAlex W ID 或 OpenAlex work URL；建议 1-10 篇。","paths":"[Truncated]","extensions":"[Truncated]","min_bytes":8,"example":"user_inputs/citation-graph-explorer/seeds.md"}],"optional_inputs":[{"id":"graph_policy","label":"图谱探索范围","description":"可选；说明最大邻居数、目标年份或语言、筛选角色、只要引用还是也要 related works。","paths":"[Truncated]","extensions":"[Truncated]","min_bytes":10,"example":"user_inputs/citation-graph-explorer/policy.md"}],"outputs":[{"id":"graph_report","label":"引文邻域报告","path":"literature/skill_citation_graph.md","description":"按种子列出实际抓取来源、引用/related 数量、候选角色、警告和建议的下一步阅读动作。"},{"id":"graph_record","label":"引文邻域结构记录","path":"literature/skill_citation_graph.json","description":"机器可读的种子、原始边、候选记录、provider fallback、抓取警告和筛选边界。"},{"id":"domain_map","label":"引文图谱领域映射","path":"literature/skill_citation_domain_map.json","description":"基于本次实际候选和引文边形成的 core/bridge/adjacent/boundary 机械映射，供人工复核。"}]} |
Citation Graph Explorer
Read and normalize each listed seed without discarding the original string. For a DOI, register the seed with process_seed_paper when possible, then call
fetch_outgoing_citations. For an OpenAlex work identifier, call
fetch_outgoing_citations directly. Respect the bounded policy; do not recursively snowball, scrape publisher pages, or silently replace an unavailable seed with a title-similar paper.
Write the raw but bounded response for every seed to skill_citation_graph.json and make every fallback or provider warning visible. Pass only the actual returned paper records and edges to build_domain_map, explicitly setting the declared
skill_citation_domain_map.json output path. Treat that map as a mechanical organization aid, never as a final gap, novelty, importance, or relevance judgment.
The Markdown report must distinguish direct references, provider-reported related works, metadata-only candidates, and sources requiring verification. It may propose section-specific follow-up searches, but it must not create final citations or assert that an unread neighbor supports a claim. Finish after all three declared output files exist, including a report of failed or unresolved seeds.
