| name | skill-endnote-research |
| description | Use when the user gives a research idea and wants Codex to analyze possible research directions, search the local EndNote litlib index, create a literature list, copy matching PDFs, and generate a literature-review folder with a report. This skill is designed for the local EndNote workspace that contains litlib/index/papers.jsonl and My EndNote Library-Saved(1).Data/PDF. |
Skill EndNote Research
Use this skill for Chinese or English research-idea prompts such as:
- "我要研究 流体与震群活动的关系。"
- "帮我从本地 EndNote 库找这个想法相关的文献并整理"
- "根据这个研究问题生成文献梳理文件夹"
Workflow
- Identify two paths and keep them separate.
- Output path: use the directory where the user invoked Codex as the project output location. Create
文献梳理/ there by default.
- EndNote/litlib workspace: use the directory that contains
litlib/index/papers.jsonl; for this user's local library, the usual path is /mnt/d/YIN/BaiduSyncdisk/endnote_file.
- If the current directory is not the EndNote workspace, still keep output in the current project directory and pass the EndNote library path via
--workspace.
- Analyze the idea before searching.
- State the likely research directions and why they matter.
- For earthquake/seismology ideas, consider directions such as earthquake swarms, repeated earthquakes, induced seismicity, fluid triggering, fault activation, catalog statistics, migration patterns, focal mechanisms, stress transfer, geothermal/reservoir/wastewater contexts, and hazard implications.
- Translate Chinese research ideas into English before searching.
- If the user prompt is Chinese, write an English translation first. Example:
我要研究 流体与震群活动的关系。 -> I want to study the relationship between fluids and earthquake swarm activity.
- Search both Chinese-language and English-language literature in the local library. Do not treat a Chinese prompt as Chinese-only retrieval.
- Generate multiple Chinese and English search queries.
- Include the original Chinese wording, the English translation, location names in Chinese and English, domain synonyms, mechanism terms, and method terms.
- For the fluid-swarm example, include queries such as
流体 震群 活动 关系, fluid earthquake swarm activity relationship, fluid triggering earthquake swarm, pore pressure seismicity swarm, and hydrothermal fluids swarm seismicity.
- Write these queries to a temporary text file when using the export script.
- Run the bundled script to create the archive.
- Script path:
scripts/export_lit_review.py
- Example:
python3 ~/.codex/skills/skill-endnote-research/scripts/export_lit_review.py \
--workspace /mnt/d/YIN/BaiduSyncdisk/endnote_file \
--question "我要研究 流体与震群活动的关系。" \
--queries-file /tmp/litlib_queries.txt \
--limit 20
- Read
文献列表.md and 检索记录.json.
- Use title/abstract matches as the primary retrieval evidence.
- Use full-text snippets for confirmation and close-reading direction.
- If the result set is weak, say so explicitly and suggest better query families.
- Write or revise
总结报告.md in the created output folder.
- Include: research-direction map, key literature groups, useful methods, initial research questions, likely data requirements, and next close-reading steps.
- Distinguish what the literature supports from Codex interpretation.
Script Behavior
The script reads:
litlib/index/papers.jsonl
litlib/index/texts/*.txt
- each record's
path field for the source PDF
It creates:
<directory where the command is run>/文献梳理/
<timestamp>-<question-slug>/
文献列表.md
总结报告.md
检索记录.json
pdfs/
Default retrieval is title/abstract first, then full-text confirmation. It copies PDFs for matched records when the source files exist.
Output Guidance
When reporting back to the user, mention:
- the created folder path
- how many literature records were selected
- how many PDFs were copied
- any missing PDFs or weak-result caveats
- the main research directions found
Do not modify litlib/litlib.py unless the user explicitly asks to change the library tool itself.