| name | pdf-note-card |
| description | Turn one uploaded or explicitly identified research paper into a source-traceable standalone paper reading note with clear locations in the paper. DOI, arXiv ID, OpenAlex ID, or a direct URL can be resolved into the declared input PDF path before reading. |
| tools | ["read_file","write_file","process_seed_paper","extract_paper_sections","extract_pdf_text","fetch_paper_pdf","fetch_paper_metadata","openalex_search","crossref_search","semantic_scholar_search","arxiv_search","finish_task"] |
| strict_tools | true |
| model_tier | standard |
| temperature | 0.15 |
| allowed_read_prefixes | ["user_inputs/pdf-note-card/","user_seeds/","literature/","_runtime/skill_sessions/"] |
| allowed_write_prefixes | ["user_seeds/","literature/"] |
| intake_tools | ["fetch_paper_pdf","fetch_paper_metadata","openalex_search","crossref_search","semantic_scholar_search","arxiv_search"] |
| outputs_expected | {"note_index":"literature/skill_pdf_note_cards/index.md","note_record":"literature/skill_pdf_note_cards/index.json"} |
| interaction | {"mode":"guided","language":"zh-CN","summary":"上传一篇 PDF，或提供 DOI、arXiv/OpenAlex ID、URL 或精确标题；系统会尝试获取 PDF，再生成带论文位置、阅读范围、方法、结果、局限和引用边界的论文阅读笔记。","request_required":false,"request_prompt":"可选：说明这篇论文的用途，例如用于 Related Work、方法比较、某个 claim，或 idea 机制分析。","example_request":"为这篇 PDF 建立中文论文阅读笔记，重点提取方法、实验、局限和可用于当前研究问题机制分析的相关内容。","required_inputs":[{"id":"paper_pdf","label":"单篇研究论文 PDF","description":"上传真实可读 PDF，或在材料准备时明确提供 DOI、arXiv/OpenAlex ID、直接 URL 或精确标题以尝试下载到此路径。Skill 会读取论文内容；扫描件、损坏文件、访问受限链接或只含封面会被明确标为无法充分阅读。","paths":"[Truncated]","extensions":"[Truncated]","min_bytes":1024,"example":"user_inputs/pdf-note-card/paper.pdf"}],"optional_inputs":[{"id":"paper_context","label":"阅读目的或已知标识符","description":"可选；粘贴 DOI、arXiv/OpenAlex ID、直接 URL、已知标题、研究问题或希望优先核验的论文内容。标识可触发受控下载尝试，但元数据或检索命中不会替代可读 PDF。","paths":"[Truncated]","extensions":"[Truncated]","min_bytes":10,"example":"user_inputs/pdf-note-card/context.md"}],"outputs":[{"id":"note_index","label":"PDF 阅读笔记索引","path":"literature/skill_pdf_note_cards/index.md","description":"指向本次动态命名的阅读笔记、源 PDF、阅读覆盖范围、证据等级和下一步建议。"},{"id":"note_record","label":"PDF 阅读笔记结构记录","path":"literature/skill_pdf_note_cards/index.json","description":"机器可读的笔记路径、来源、阅读覆盖范围、引用边界和未解决问题。"}]} |
PDF Note Card
Read the intake packet and the resolved or uploaded PDF first. When the intake record lists a DOI/arXiv/OpenAlex ID or URL, preserve that identifier and its download outcome in the note provenance. Call process_seed_paper with
source="pdf_path" and value="user_inputs/pdf-note-card/paper.pdf" so the original source is registered under user_seeds/ without overwriting another paper. Then call
extract_paper_sections for the available core sections. Use extract_pdf_text only when a required fact is absent from the bounded section extraction, and preserve the tool's page-coverage/truncation information.
Create a dynamically named Markdown note beneath literature/skill_pdf_note_cards/ and update both declared index files. The card must separately record: source path and resolved metadata; pages/sections actually read; FULL-TEXT, PARTIAL-TEXT, or
UNREADABLE evidence status; problem; method; data/evaluation; reported results;
limitations; mechanism/design rationale when present; exact section anchors; allowed citation use; and unresolved facts. A source PDF does not justify an invented DOI, BibTeX entry, numerical result, or causal claim.
This is a standalone note-card lane, not a substitute for the canonical T3 queue. Do not write literature/deep_read_notes/ or claim a T3 deep-read note is complete. If the PDF cannot support the requested extraction, write the index with the concrete failure and use ask_human for a readable PDF, DOI, or clarification. Finish by naming the dynamic note-card path and both declared index paths.
