All products

Check research sources with AI

Keep claims tied to inspected sources and a search log. Local scripts check citation structure, identifiers and reachability, leaving substantive support for human review.

Check research sources with AI when a report needs more than a plausible reference list. The research-integrity workflow defines the question, records searches, compares evidence and connects claims to sources. A search result or abstract must remain distinct from a source whose full text was actually inspected.

Local scripts check numbering, dates, URLs, titles and article identifiers. Named NCBI and Crossref services provide structured metadata checks. Those results cannot establish that a passage supports the report's conclusion. Substantive claims still need comparison with the original text, while unresolved access and conflicting evidence remain visible in the delivered report.

What it does

  • Keep a reproducible search log.
  • Check citation numbering and required fields.
  • Compare titles with authoritative identifier metadata.
  • Flag unsupported or unresolved claims.
  • Distinguish independent evidence from repeated publicity.
  • Retain limitations beside the conclusions they affect.

Install and use

  1. Install the full package in a web- and file-capable agent with Python for local checks. Search/browser tools and optional AMiner access must be separately configured. Supply only authorised reports and source material.
  2. Ask, for example, “Review this report on paper recycling. Check its citations and numerical claims, and identify sources available only as abstracts.”
  3. Read the check results, distinguishing a broken identifier from a blocked access channel. A blocked response does not justify inventing the missing text.
  4. Compare important claims with their actual sources and retain unresolved findings in the final report.

FAQ

Source

Made by Liozhang. Original skill: research-integrity, published on GitHub under the MIT licence.

View on GitHub