Research Hound is a scientific research workflow for Claude Code. It runs five stages in order (critical thinking, literature review, brainstorming, writing and evaluation) and backs them with 21 runnable scripts that verify what a chat model would otherwise just assert.
- Citation verification:
verify_citations.pychecks that each reference's DOI resolves. - Claim audits: every claim in a manuscript is traced to its evidence.
- Manuscript linting: an IMRaD scaffold, consistency and reference checks, and CONSORT, PRISMA or STROBE selection.
- Rubric scoring with inter-rater agreement.
Only the citation check touches the network. No script needs an API key, and nothing sends your prose, data or figures to a hosted model. Python 3.10+. Part of a three-skill suite with Skill Vision and Agent Oracle.
Read more: Claude AI skills for research and agent design.