radiology-skills plugin
Forty skills for Claude and Codex on imaging research: study design, radiomics, statistics and journal submission.
- 1,981 ★ on the repository
- Last commit
Extended access
Why this level
- Skills run scripts from their directories and work with research data that may contain patient information.
Install
- Works in
Manual install
mkdir -p "$HOME/.codex/skills"
for skill_dir in radiology-skills/skills/radiology-*; do
cp -R "$skill_dir" "$HOME/.codex/skills/"
doneRun from the root of a repository clone (macOS and Linux). `install.md` has a PowerShell variant for Windows and separate Claude instructions in `radiology-skills/install.md`.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Skills run scripts from their directories and work with research data that may contain patient information.
This is third-party code. Review the repository files before installing.
What it does
The package bundles forty independent `radiology-*` skills that guide medical imaging research from framing the question to publication. The benchmark is the requirements of the journal Radiology and the Nature family. Modules cover quality control and reconstruction, annotation, deep learning, radiomics and radiogenomics, statistics, systematic reviews, ethics and reproducibility. Separate skills work with texts: `radiology-writing` and `radiology-polishing` prepare a manuscript, `radiology-prereview` runs a pre-submission review, `radiology-response` answers reviewers, `radiology-grant` checks a grant application, and `radiology-figure` and `radiology-table` lay out figures and tables. `radiology-pipeline` is meant for multi-stage tasks. Each skill first determines what claim the study makes, then picks suitable reporting standards such as CLAIM, TRIPOD+AI, STARD-AI, CLEAR and IBSI. The skill distinguishes evidence states: provided, executed, verified, missing and conflicting.
Who it is for. Radiology research groups working with CT, MRI, PET or ultrasound data who prepare radiomics and AI papers and grant applications.
Good fit when
- Multicenter MRI data is collected and it is unclear which study can be built from it.
- A radiomics pipeline is checked for data leakage between training and test sets.
- A manuscript is read before journal submission the way a reviewer would read it.
- A reply to reviewer comments is assembled point by point with evidence.
- A grant application is checked against the current year's requirements.
Not a fit when
- A clinical diagnosis from a scan is needed: the package helps with research methodology, not treatment.
- Guaranteed acceptance of a paper or grant is expected: the author makes no such promise.
- The work is unrelated to medical imaging and forty modules would be excessive.
Example request
Use radiology-radiomics to check preprocessing, feature selection and validation for data leakage in our liver cancer MRI study.Limitations
Documentation and some rules are written in Chinese, while the output can be Chinese, English or mixed. The author states that structural checks pass but there is no independent evaluation of model performance, and promises no acceptance of a paper or grant. The agent must check journal and funder requirements against current official rules each time. Installation copies the `radiology-*` directories whole, together with `references`, `scripts` and `agents`. The research team verifies data, analysis, ethics approval and conclusions.
How to disable. Delete the copied `radiology-*` directories from `~/.codex/skills` or from the Claude skills folder and restart the agent. An older version with the same name must be removed before installing a new one.
FAQ
How do I install the package in Codex?
Run `git clone https://github.com/huang-sir1/radiology-skills.git`, enter the directory and on macOS or Linux run `mkdir -p "$HOME/.codex/skills"`, then copy each `radiology-skills/skills/radiology-*` into it. Restart Codex.
How do I install it in Claude?
Details on the local plugin and single-skill install are in `radiology-skills/install.md`. The inner `radiology-skills/` directory holds `.claude-plugin` and `.codex-plugin`. The author does not claim a remote marketplace listing.
Where do I start a multi-stage job?
Call `radiology-pipeline`: it identifies the research question, existing evidence and missing materials, then proposes the next tasks.
Which reporting standards does the package use?
CLAIM, TRIPOD+AI, STARD-AI, CLEAR and IBSI, chosen by study type. Reporting completeness and methodological validity are reviewed separately.
Does the package guarantee publication?
No. The author promises no acceptance of a paper or grant and claims no independent evaluation of model performance.
Discussion
No questions or reviews yet.
Related
PluginsSkills for clinical research: literature search, statistics, reporting guideline and citation checks, journal submission
Research Skills
research-skills
Skills for scientific work: literature reviews on AI in medical imaging, a manuscript audit, slides. No paywall bypass, data comes from OpenAlex and PubMed
Skills tailored to specific academic journals that help choose a venue, prepare a paper and pass review
Ten research skills for Claude Code and Codex: figures, conference and journal papers, grants, peer review and code release
A library of 280+ research skills: literature search, experiment design, analysis, reference checks and paper preparation
Research Skills gives Claude Code, Codex and Copilot CLI plugins for research: literature, grants, papers, figures, slides.
README in short
The package is made for Chinese research groups aiming at journals of the Radiology and Nature level. The forty modules are framed as virtual specialist consultants, each called for the task at hand.