research-paper-notes
Skill8 starsChanged 5 months agoMITView on GitHub
Build topic-centered Markdown research-paper reading notes with a survey overview and per-paper deep notes. Use when the user asks to create, expand, or restructure paper-reading notes from: (1) specific paper titles, PDFs, or links, (2) a research topic, or (3) the user's own paper from which the agent should infer a related topic. The workflow emphasizes paper selection, source-backed reading, method and experiment explanation, and especially writing/positioning analysis that contrasts the paper's plain underlying contribution with the authors' packaged story.
In this repository58 other files it gives its agents.
- skills/writing-polish/humanizer/AGENTS.mdAGENTS.md
- frontend-designSkill
- ask-mattSkill
- codebase-designSkill
- code-reviewSkill
- diagnosing-bugsSkill
- domain-modelingSkill
- github-codex-keySkill
- github-ssh-key-pushSkill
- implementSkill
- improve-codebase-architectureSkill
- prototypeSkill
- setup-matt-pocock-skillsSkill
- tddSkill
- to-issuesSkill
- to-prdSkill
- triageSkill
- huggingface-accelerateSkill
- quantizing-models-bitsandbytesSkill
- gpuboxSkill
- evaluating-llms-harnessSkill
- ml-training-recipesSkill
- peft-fine-tuningSkill
- experiment-tracking-swanlabSkill
- unslothSkill
- matlab-add-awgnSkill
- matlab-agentic-toolkit-setupSkill
- matlab-build-appSkill
- matlab-create-live-scriptSkill
- matlab-debuggingSkill
- matlab-design-digital-filterSkill
- matlab-display-imageSkill
- matlab-install-productsSkill
- matlab-list-productsSkill
- matlab-map-database-objectsSkill
- matlab-model-serdes-systemsSkill
- matlab-modernize-codeSkill
- matlab-read-databaseSkill
- matlab-review-codeSkill
- matlab-testingSkill
- matlab-use-duckdbSkill
- matlab-write-databaseSkill
- grillingSkill
- grill-meSkill
- grill-with-docsSkill
- handoffSkill
- teachSkill
- writing-great-skillsSkill
- ara-compilerSkill
- ara-research-managerSkill
- ara-rigor-reviewerSkill
- brainstorming-research-ideasSkill
- creative-thinking-for-researchSkill
- academic-plottingSkill
- ml-paper-writingSkill
- paper-spineSkill
- humanizerSkill
- prompt-optimizerSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
24 rules. About 0 are common, 19 are distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈19
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 10
Create a reusable research-reading note set, not a loose summary.
- Distinctive: line 17
Write generated notes in the user's requested language.
- Distinctive: line 25
Specific papers: use provided titles, PDFs, DOIs, arXiv links, conference links, or local files as the paper set.
- Distinctive: line 26
Topic/topic phrase: search for representative papers, then ask the user to confirm the candidate list before writing full notes.
- Distinctive: line 27
User's own paper: read the user's manuscript first, infer the closest research topic and comparison space, then search and propose …
- Distinctive: line 29
If the user provides no retrieval constraints, choose a defensible scope: recent top-tier or high-impact work, plus a few foundational or …
- Very specific: line 37
Read references/search-selection.md before searching or proposing a paper list.
- Distinctive: line 42
Collect sources: prefer primary sources and stable metadata, including conference proceedings, ACL Anthology, OpenReview, arXiv, …
- Distinctive: line 44
Confirm the list: ask the user to approve, remove, add, or reorder papers.
- Very specific: line 45
Choose and announce the output root: follow references/workspace-layout.md to choose the note library root.
- Very specific: line 49
Write the overview note: create a survey-style map that links to each per-paper note.
- Distinctive: line 50
Write per-paper notes: every note must be detailed.
- Distinctive: line 51
Verify navigation and completeness: check that overview links, per-paper links, PDFs, numbering, required overview diagrams, and note …
- Very specific: line 53
Read references/note-schema.md before writing the overview or per-paper notes.
- Very specific: line 54
Read references/workspace-layout.md before creating or reorganizing files.
- Distinctive: line 58
Every per-paper note must explain:
- Distinctive: line 70
Use short original phrases or sentences only where they teach writing or framing.
- Distinctive: line 74
Prefer this structure unless the existing workspace has a stronger convention:
- Distinctive: line 102
Prefer accuracy over speed.
- Distinctive: line 103
Use primary sources for technical and publication claims whenever possible.
- Distinctive: line 104
Do not invent code links, datasets, numbers, or experimental results.
- Distinctive: line 105
If a PDF cannot be downloaded, list the exact title and ask the user to provide it.
- Distinctive: line 106
If the note set already exists, extend and refactor it in place instead of creating a disconnected duplicate.
- Distinctive: line 107
Keep note headings formal and research-oriented.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Research Paper Notes
Core Promise
Create a reusable research-reading note set, not a loose summary. The output should help the user:
- choose research topics;
- learn paper writing, framing, and packaging;
- understand methods and experiments deeply enough to discuss the paper;
- navigate from one overview note to detailed per-paper notes.
Write generated notes in the user's requested language. If the user does not specify a language, use the language of the user's task request.
PDF export is out of scope unless the user separately requests it.
Input Modes
Handle three entry points:
- Specific papers: use provided titles, PDFs, DOIs, arXiv links, conference links, or local files as the paper set.
- Topic/topic phrase: search for representative papers, then ask the user to confirm the candidate list before writing full notes.
- User's own paper: read the user's manuscript first, infer the closest research topic and comparison space, then search and propose candidate papers for confirmation.
If the user provides no retrieval constraints, choose a defensible scope: recent top-tier or high-impact work, plus a few foundational or highly relevant papers when needed for context.
Mandatory Gate
For topic-based or user-paper-based tasks, do not generate the full note library immediately. First produce a candidate paper list with titles, venues/years, links, categories, and selection rationale. Wait for user confirmation unless the user explicitly says to proceed without review.
For a fixed paper set, proceed directly unless paper identity is ambiguous.
Read references/search-selection.md before searching or proposing a paper list.
Workflow
- Clarify scope only when necessary: infer reasonable defaults for venue tier, years, and paper count if the user did not specify them.
- Collect sources: prefer primary sources and stable metadata, including conference proceedings, ACL Anthology, OpenReview, arXiv, DOI/publisher pages, official project pages, and author-released code/data pages.
- Create a candidate list: categorize papers by research role rather than only chronology.
- Confirm the list: ask the user to approve, remove, add, or reorder papers.
- Choose and announce the output root: follow
references/workspace-layout.mdto choose the note library root. Do not treat any example path as hard-coded. State the output root before writing files. - Organize files: save PDFs and notes under that topic workspace. Preserve existing user files and do not overwrite unrelated notes.
- Read each paper: extract text from the PDF when possible; use the original PDF for method, experiment, table, and writing claims.
- Use subagents when supported: for full note libraries, delegate category-level reading/synthesis passes to subagents or parallel worker agents when the host system supports them. Assign each worker one research category or method family, with all papers in that category. Require each worker to produce a category synthesis plus source-grounded per-paper reading notes for method, experiments, limitations, and packaging moves. If subagents are unavailable, read categories sequentially and state that fallback.
- Write the overview note: create a survey-style map that links to each per-paper note. For multi-paper libraries, include the required Mermaid research storyline and method taxonomy diagrams from
references/note-schema.md. - Write per-paper notes: every note must be detailed; do not split content into "brief" and "deep reading" sections.
- Verify navigation and completeness: check that overview links, per-paper links, PDFs, numbering, required overview diagrams, and note sections are consistent.
Read references/note-schema.md before writing the overview or per-paper notes.
Read references/workspace-layout.md before creating or reorganizing files.
Quality Bar
Every per-paper note must explain:
- what problem the authors claim to solve;
- what the paper actually does after removing terminology and packaging;
- how the method works step by step;
- how experiments are organized, what each experiment is meant to prove, and what the results do or do not support;
- how the authors package the contribution into a stronger paper story;
- what short original expressions are worth learning for writing;
- what limitations, open questions, and future research ideas follow from the paper.
The packaging section is mandatory. Always include a contrast between plain essence before packaging and author-packaged story after packaging.
Use short original phrases or sentences only where they teach writing or framing. Do not turn notes into long copied excerpts.
Output Shape
Prefer this structure unless the existing workspace has a stronger convention:
papers/
01_category_name/
02_category_name/
notes/
README.md
papers/
01_slug.md
02_slug.md
.paper_text/The overview note should include:
- topic-level thesis;
- classification/storage map;
- paper table with links;
- Mermaid research storyline and method taxonomy diagrams for multi-paper libraries;
- cross-paper comparison tables;
- writing and packaging pattern summary;
- research direction candidates.
Per-paper notes should link back to the overview and to the local/source PDF when available.
Working Rules
- Prefer accuracy over speed. If paper metadata, publication status, or venue may have changed, verify online.
- Use primary sources for technical and publication claims whenever possible.
- Do not invent code links, datasets, numbers, or experimental results.
- If a PDF cannot be downloaded, list the exact title and ask the user to provide it.
- If the note set already exists, extend and refactor it in place instead of creating a disconnected duplicate.
- Keep note headings formal and research-oriented; avoid casual headings such as "one-sentence summary."
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

