literature
Skill3 starsChanged 15 days agoMITView on GitHub
Reading, judging and citing papers in an acl-paper project (has paper.config.json). Use before citing a paper, when deciding whether a paper is relevant, when checking that a citation supports its claim, when the user drops a PDF into the project or pastes a BibTeX entry, and when the bib sync reports unknown or unread keys.
5 other projects use a version of this.
Adopted by
Most recently taken up by
- LeoLourson/MathPaperLibtheir own version, first seen Nov 15, 2026
- sghng/abstracttheir own version, first seen Oct 15, 2026
- MercatorProj/vibe-thesis-skillstheir own version, first seen Oct 10, 2026
- wu-yc/LabClawtheir own version, first seen Oct 8, 2026
- hsigstad/research-kittheir own version, first seen Jul 5, 2026
Other versions5 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 5
- wu-yc/LabClaw★ 1.1ksame name
- MercatorProj/vibe-thesis-skills★ 8same name
- LeoLourson/MathPaperLib★ 3same name
- hsigstad/research-kit★ 0same name
- sghng/abstract★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository13 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- templates/CLAUDE.mdCLAUDE.md
- acl-anthologySkill
- arxivSkill
- camera-readySkill
- guidelines-checkSkill
- initSkill
- new-sourceSkill
- proofreadSkill
- rebuttalSkill
- reviewerSkill
- revisionSkill
- writingSkill
The file
We split this skill into the rules it gives an agent
4 rules. 0 are common, 2 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 4 edited versions we read keep these.
2
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 21
Run PK mark-read <key>. It refuses while Read or Claims is empty or the .md is missing.
- Distinctive: line 23
When you tell the user a paper supports a claim, quote the body passage in chat.
- Very specific: line 28
Keep one key per paper. Before citing, check literature/index.md for the same paper under another key (preprint and published version).
- Distinctive: line 29
After a rewrite that touches a cited sentence, each resulting sentence that rests on the source keeps its citation.
“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
How papers move from "found" to "cited" in this project. Searching and fetching are the source skills' job (acl-anthology, arxiv, project sources in .claude/skills/); everything after the fetch is here.
PK means python3 ${CLAUDE_PLUGIN_ROOT}/scripts/paperkit.py, run from the project root (the folder with paper.config.json).
The rule
A paper is judged and cited only on its body: introduction, method, experiments, results, tables, limitations. The abstract (and a search hit's snippet) is for triage: it decides what to read, never what a paper shows. A citation stands once the body passage that supports it is in the notes.
Reading a paper
- Open
literature/md/<key>.md. For long papers, grep it for the terms of the claim and the section headers first, then read the matching sections in full. Use the PDF (literature/pdf/<key>.pdf) for figures, table layout and printed page numbers. - Fill
literature/notes/<key>.md:## Read: which parts you read ("full paper", "Sec. 1-5, Tab. 2").## Claims this paper supports: one line per claim, in your words, with its body location and a short quote. Numbers exactly as printed in the paper.## Relevance to our paper: where and why it would be cited, or why not.
- Run
PK mark-read <key>. It refuses whileReadorClaimsis empty or the.mdis missing; fix what it names.
When you tell the user a paper supports a claim, quote the body passage in chat.
Citing
- Cite only
read-fullkeys, with\citet{}/\citep{}. Write the citation into the.texand stop there: the PostToolUse hook moves the entry fromliterature/archive.bibinto the paper bib, and moves it back when the last citation goes. - Keep one key per paper. Before citing, check
literature/index.mdfor the same paper under another key (preprint and published version); when both exist, ask the user which to cite. - After a rewrite that touches a cited sentence, each resulting sentence that rests on the source keeps its citation.
Entries and PDFs that do not come from a source skill
- User pastes a BibTeX entry: save it to a temp file and run
PK bib-paste --file <tmp>. - User drops a PDF into the project: identify the paper from its first page, find its key in the paper bib or
literature/archive.bib, and runPK add-local <pdf> --key <key>. No entry yet: stop with ⚠️, name the paper (title, authors, year, venue or DOI) and ask the user for the authoritative entry. File the PDF once the entry exists. - A paper with no fetchable source (book, report): ask for the entry and a PDF, then the two steps above.
Sync messages
The hook (and PK sync) reports:
- unknown key: cited but in neither bib file. Fetch it through a source skill, or ask the user for the entry. Write no BibTeX yourself.
- not read in full: cited before its notes were done. Read it now (steps above). If the body does not support the sentence, tell the user and propose a fix.
PK index rebuilds literature/index.md; PK sync runs everything by hand.
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.

