Repository navigation
Errors and Edge Cases
Most social-science journals are not in PubMed, and a closed paper has no open copy. The script says both and the review is graded on the abstract, which it says in the verdict:
$ python3 paper_meta.py 10.1111/jssr.70101 --similar 0 | grep -v '^Abstract'
== CROSSREF
Title: Cognitive Ability and Christian Nationalism
Journal: Journal for the Scientific Study of Religion | Publisher: Wiley | Type: journal-article
DOI: 10.1111/jssr.70101 | Published: 2026-09-11 | Online: 2026-09-11
received: 2026-06-07
accepted: 2026-08-21
published: 2026-09-11
Authors:
- Samuel L. Perry [Department of Sociology University of Oklahoma Norman Oklahoma USA]
License: http://onlinelibrary.wiley.com/termsAndConditions#vor
References deposited: 77
== UNPAYWALL
skipped: no contact email (pass --email or set SSE_EMAIL)
== PUBMED
no PMID (too new, or the journal is not indexed in PubMed)
== NEXT
Open-copy check skipped (no email): rerun with --email or SSE_EMAIL, or try the publisher page; if it stays closed, grade at abstract level and say so.
$ echo $?
0
Here the open-copy check was skipped because no email was set (Configuration); set one and the UNPAYWALL section runs.
Crossref carries retraction notices from the publisher and, since January 2025, from Retraction Watch. The script prints each as a !! line and puts an integrity notice first in NEXT:
$ python3 paper_meta.py 10.1057/s41599-025-04787-y --similar 0 --email you@example.org | grep -E '^(Title|!!|== |INTEGRITY|PMID)'
== CROSSREF
Title: RETRACTED ARTICLE: The effect of ChatGPT on students’ learning performance, learning perception, and higher-order thinking: insights from a meta-analysis
!! updated by: retraction 10.1057/s41599-026-07310-z (publisher 2026-04-22)
!! updated by: retraction 10.1057/s41599-026-07310-z (retraction-watch 2026-04-22)
== UNPAYWALL
== PUBMED
== NEXT
INTEGRITY NOTICE on this record: read it before anything else.
$ echo $?
0
That run keeps only the headers and notices (grep -E). PubMed's own retraction and erratum links print as !! lines in the PUBMED section when it has them (from the source and the unit tests; this paper has no PubMed record).
$ python3 paper_meta.py "not a doi"
error: no DOI, PMID or PMCID found in 'not a doi'
$ echo $?
2
Several sites refuse automated fetches, and the skill routes around them rather than retrying. The full table, with dates, is references/fetch-routes.md:
| Site | What happens | What the skill does |
|---|---|---|
| Wiley, SAGE | a "verify you are human" check | Crossref and Europe PMC, then one browser try, then abstract level |
| Elsevier journals | 403 to scripted requests | Europe PMC once the paper is a few months old, else a real browser |
| pmc.ncbi.nlm.nih.gov | a reCAPTCHA page | Europe PMC's full-text endpoint for the same PMC number |
| clinicaltrials.gov | robots.txt refuses fetch tools | the trial's published protocol paper |
| reddit.com | blocked | searches the post's title and reads the outlet's article |
The script never requests a publisher site, so these blocks don't affect it.
In a hosted app whose sandbox blocks outgoing requests, the skill uses the agent's own web tools on the same APIs one at a time, and reads PubMed's web page in a browser. Not tested in 0.1.0.
--help names every option but describes only identifier and --version; the descriptions are in the script's docstring and on Commands. A gap in 0.1.0.
PubMed's list is ranked by PubMed. In the Commands run, its first entry for a sleep-and-memory paper was "Personality Theories". The skill scans the list for a companion paper or a contradicting design and ignores the rest.
Crossref, PubMed and Unpaywall records are updated: a paper gets a PMC copy months after publication, a retraction notice appears, an abstract is added. Rerun the script for the status today rather than trusting an old run.
This wiki describes science-study-evaluator 0.1.0 and was last updated on 2026-10-11. The skill and script are MIT licensed. Report problems in the issues.
Using it
The releases
Contributing
Elsewhere