Repository navigation
FAQ
No. It says how well a study supports its claims. On a health topic the review reminds you that trials, not mechanism debates, settle which treatments work; talk to a clinician before changing a treatment.
The full text was closed and no open copy existed, so the agent graded what it could reach and said so. It names the one number to look for if you get the PDF. See Errors and edge cases.
A good journal is one check of many. The reviews that shaped the checklist found a top-journal paper whose headline rested on hit rate while d′, the measure of memory itself, showed no difference. Another headlined an equivalence claim made from p > 0.05 with no margin. The venue is graded; it doesn't decide the verdict.
Yes. It reads the paper and the public databases. The script needs outgoing HTTPS to the four hosts in Configuration.
Unpaywall's terms ask for a contact address on each request, and NCBI asks tools to identify themselves. The script sends nothing else about you. Without an email it skips Unpaywall and still runs.
In your folder: SSE_HOME, or ~/.science-study-evaluator/ (Configuration). The skill reads the last lines of the verdict log before each review.
Yes: edit references/methodology.md in your copy and bump its version line. To share a change, open a pull request with the case that showed the gap; most checklist items came from a real review the old checklist missed.
The skill follows the Evergreen Protocol. About every 30 days it searches primary sources (EQUATOR, Cochrane, GRADE, Retraction Watch, Crossref) for changes to reporting guidelines, bias tools and integrity screening, and logs each finding with its date. Its next check is due 2026-10-30. See Development.
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