download
Skill2 starsChanged 2 months agoMITView on GitHub
Add papers to the Zotero library from links or PDFs (fetch, classify, save, verify, log). Use when the user gives paper links or asks to add / download / save a paper.
- Reads credentials
2 other projects use a version of this.
Adopted by
Most recently taken up by
- jantichy/claudetheir own version, first seen Nov 1, 2026
- timelapsetech/mediaskillstheir own version, first seen Aug 15, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- jantichy/claude★ 52same name
- timelapsetech/mediaskills★ 1same 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.
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 19
Run (the command waits up to 150 s per paper for the server sync — allow a 10-minute tool timeout, and don't poll):
- Distinctive: line 30
Read the output and act only where needed (usually nowhere):
- Very specific: line 59
Look before saving: python3 zl.py fetch <link> (full card) -> python3 zl.py suggest <slug> -> python3 zl.py save <slug> ....
- Distinctive: line 64
Write to sqlite, or upload PDFs through the Web API (attachment sync is WebDAV.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
download — add papers to Zotero
The judgement is done by the scripts (taxonomy.toml rules + the local model, or you when ZC_LLM=agent); you run one
command, relay its report, and handle only what it marks. Never read the PDF or grep the full text yourself.
Arguments: one or more links or local PDF paths, space-separated (arXiv / DOI / OpenReview / PDF / project page / a
paper page with citation meta such as JMLR / a publisher page with the DOI in its path). A paper indexed nowhere yet (a
project page with a camera-ready PDF, no arXiv, no DOI) is still added: the local model reads title, authors, abstract and
the venue statement off the PDF's first page, the date is the PDF file's creation date — the report flags all of it. A file with one link per line (a reading list, a bibliography): --from-file FILE
instead of the links. Ask which papers if none were given. On Windows use python instead of python3.
Steps
Run (the command waits up to 150 s per paper for the server sync — allow a 10-minute tool timeout, and don't poll):
python3 zl.py add <links...>For each paper it fetches metadata + PDF + full text, checks for duplicates, classifies (collection + tags with evidence), saves through the Zotero desktop app, waits for the cloud sync and verifies, commits the audit log, and finally prints a
| key | title | collection | tags | PDF | notes |table.- Read the output and act only where needed (usually nowhere):
- saved: glance at the evidence lines under "assign"; an obviously wrong tag (evidence is related work or a
baseline) ->
python3 zl.py untag <key> <tag> --why <reason>. Otherwise leave it. - candidates (listed, not assigned): put them in the report for the user to decide; one whose evidence clearly
meets the definition in
taxonomy.tomlmay be added right away withpython3 zl.py tag <key> a,b— say so. - PAUSE (not saved; the exact command to finish is printed):
- collection undecided: read the printed abstract, decide the collection, run the printed
python3 zl.py save <slug> --collection "..."line (--drop a,bremoves suggested tags). - adjudication pending (
ZC_LLM=agent, an ADJUDICATE block): for each listed candidate decide from its definition, the evidence snippets and the setup excerpt whether the paper's own method or experiments meet the definition (a baseline, related work or a data-collection rig does not count). Run the printed line with--confirm tag1,tag2for the ones that do,--confirm noneotherwise. Several papers -> onesaveline each.
- collection undecided: read the printed abstract, decide the collection, run the printed
- duplicate: skipped by default; say which item already exists.
--forceonly if the user insists. - x (error): the link was not recognized or metadata failed; tell the user why and ask for an arXiv/DOI link, the paper page, or the PDF.
- metadata from the pdf first page (not indexed anywhere): relay the title and authors for the user to check and
say the date is the file's creation date;
python3 zl.py set <key> title=...fixes what the user corrects. Don't search the web for the paper — the script already tried arXiv by title. - venue still
arXiv <- default: the script already checked the arXiv comment, PDF first page, Semantic Scholar, Crossref and the project page — don't search again; note "no acceptance found". If the user asks, search the web and fix withpython3 zl.py set <key> title="[date] [Venue] Title". - sync unconfirmed: the Zotero client hasn't uploaded the item yet;
python3 zl.py verify <key>later. No polling. - The user says "read first": add
--first(status:to-read-first).
- saved: glance at the evidence lines under "assign"; an obviously wrong tag (evidence is related work or a
baseline) ->
- Report: relay the printed table as is (don't rewrite it), then at most two or three sentences: candidates to confirm, flags that need the user's decision. The log and git are already handled by the script.
Related commands
- Look before saving:
python3 zl.py fetch <link>(full card) ->python3 zl.py suggest <slug>->python3 zl.py save <slug> .... - Re-verify existing arXiv items:
python3 zl.py recheck [--dates] [--search];--writeafter approval.
Don't
- Write to sqlite, or upload PDFs through the Web API (attachment sync is WebDAV; the clients would never see them).
- Change titles / tags of existing items on your own, create collections, or invent tags outside
taxonomy.toml. - Print the API key from
.env.
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.

