| name | conference-papers-readsimply |
| description | Use when Codex needs to generate abstract-only Obsidian Markdown notes from one or more data/papers/*.json files by reading arXiv HTML abstract metadata when available, without reading the full paper. Trigger for requests like "Use $conference-papers-readsimply 简读 data/papers/xxx.json", "批量简读 data/papers/*.json", or "generate simple abstract md notes"; this skill is independent from website note JSON and dist build. |
Conference Papers ReadSimply
Generate standalone Obsidian Markdown notes from data/papers/*.json using abstract only. Prefer arXiv HTML for the abstract and page metadata when available, but stop at the abstract block. This is a lightweight Markdown export workflow: do not edit paper["note"], do not modify data/papers/*.json, do not run website validation/build scripts, and do not treat the result as site data.
This skill is adapted from conference-papers-read2md, but it intentionally does less work to save time and tokens.
Workflow
- Resolve the site root and target paper JSON input.
- Accept a single JSON path, multiple JSON paths, a paper id/title substring, a glob such as
data/papers/*.json, or a directory such as data/papers/.
- For one paper, run:
python3 scripts/paper_json_context.py data/papers/PAPER_ID.json
python3 scripts/batch_paper_json_context.py data/papers/*.json
python3 scripts/batch_paper_json_context.py data/papers/
-
Read the helper JSON output.
output_dir: normally data/_inbox/{paper-id}
simple_note_path: normally data/_inbox/{paper-id}/{MethodName}-simple.md (for example, MethodName-simple.md)
primary_source: should be arxiv_html_abstract when an arXiv HTML URL is available.
- use
suggested_method_name as a starting point, but correct it if the title clearly uses a different method/model name.
-
Read abstract only.
- 只阅读 abstract;this skill is for abstract-only reading.
- Read arXiv HTML abstract metadata first when
arxiv_html_url is available.
- 只读取 arXiv HTML 的摘要和元信息:title, authors, date, and abstract.
- Use the helper below when possible:
python3 scripts/arxiv_html_metadata.py "$arxiv_html_url"
- Combine arXiv HTML metadata with paper JSON metadata: use arXiv HTML for title/authors/date/abstract when it is more complete, and use paper JSON for venue, year, topics, project/code links, and output paths.
- If arXiv HTML is unavailable or has no abstract block, fall back to the paper JSON
abstract, arXiv abs/API metadata, or a conference detail page abstract.
- Do not read arXiv HTML body after the abstract.
- Do not download or parse PDFs.
- Do not read method, experiment, appendix, figures, tables, formulas, or project page body content.
- If no abstract is available, stop that paper and record a clear failure.
-
Write the Markdown note using assets/paper-note-template.md.
- Fill YAML frontmatter with paper JSON metadata where available, refined by arXiv HTML metadata when available.
- Fill the metadata table with authors, venue as
{Venue} {Year}, categories/topics from the paper JSON, and dates/links from the best available metadata.
- Write one concise Chinese sentence summarizing the abstract. Do not invent method details beyond the abstract.
- For batch mode, process papers one by one. A failure for one paper must not block other papers; report success and failure lists at the end.
-
Keep the note lightweight.
- Output only frontmatter, title,
## 元信息, and ## 一句话总结.
- Do not create
assets/.
- Do not download images.
- Do not create concept notes or Obsidian MOCs.
Quality Requirements
- abstract only: never spend tokens on full-paper reading for this skill.
- Prefer arXiv HTML abstract metadata when available, but never continue into Introduction, Method, Experiments, Appendix, figures, tables, or formulas.
- Preserve only metadata and one-sentence summary.
- The summary must be based on the extracted abstract and should stay under 50 Chinese characters when possible.
- Batch output should use
{MethodName}-simple.md, overwriting existing simple notes if regenerated, but never overwriting {MethodName}.md deep-reading notes.
- Never write a deep-reading skeleton. If the abstract is missing, report the failure instead of filling placeholders.
Important Non-Goals
- Do not modify
data/papers/*.json.
- Do not generate or validate
paper["note"].
- Do not call
../conference-papers/scripts/build_site.py, validate_notes.py, or resolve_links.py.
- Do not refresh Obsidian MOCs or run git automation.
- Do not read the full paper as a fallback.
Bundled Resources
assets/paper-note-template.md: abstract-only Obsidian note template.
scripts/paper_json_context.py: normalize one paper JSON into readsimply source/output context.
scripts/batch_paper_json_context.py: expand JSON paths, globs, directories, ids, or title matches into a best-effort batch context.
scripts/arxiv_html_metadata.py: optional helper to extract title, authors, date, and abstract from arXiv HTML only.