Calkit plugin
A skill plugin for Calkit projects that builds a reproducible research pipeline and traces where manuscript numbers came from.
- 62 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The agent runs the pipeline and `calkit` commands, which execute project scripts and containers.
- Linking Overleaf and remote storage sends project data to external services.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/calkitDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The agent runs the pipeline and `calkit` commands, which execute project scripts and containers.
- Linking Overleaf and remote storage sends project data to external services.
This is third-party code. Review the repository files before installing.
What it does
Calkit joins Git, DVC, environments and a pipeline into a research project that rebuilds with one command from raw data to the paper. The `agent-plugin` folder in the repository holds a skill set that teaches an agent to work with such a project. The base `conventions` skill explains the `calkit.yaml` structure, environments, pipeline stages and CLI commands. The other skills cover single tasks. `create-pipeline` turns an ordinary folder of scripts into a pipeline, `add-pipeline-stage` adds a stage, `build-paper-pipeline` builds stages for a paper, `check-reproducibility` finds manuscript numbers with no provenance, `check-questions` compares research questions against their evidence, `connect-overleaf-project` links an Overleaf project, and `resolve-merge-conflicts` untangles merge conflicts.
Who it is for. Researchers and graduate students who keep calculations and the paper in one repository and want an agent that does not break reproducibility.
Good fit when
- A folder of scripts and data must become a project that builds with `calkit run`.
- A manuscript contains a number and it is unclear which pipeline stage produced it.
- A stage with a notebook or a new environment is added to a finished pipeline.
- The manuscript lives in Overleaf and the calculations in Git, and they must be linked.
Not a fit when
- The project does not use Calkit and has no plan to switch.
- You need a general build system with no tie to Git and DVC.
- The agent is allowed to save figures and notebooks directly: the plugin rule forbids such artifacts outside the pipeline.
Example request
Turn this folder into a Calkit project and show which pipeline stages you would add before writing any files.Limitations
The skills only work inside a project with a `calkit.yaml` file, and the `calkit` CLI installs separately through `uv tool install calkit-python` or the installer script. The agent writes code and the pipeline description but never saves computed results: figures, executed notebooks and PDFs must come from a `calkit run`. Heavy stages may need an HPC cluster and configured Docker, Conda, R, Julia or MATLAB environments. For Codex and OpenCode the skills are installed with `calkit update agent-skills`, and auto-update of the Claude Code plugin is enabled in the marketplace settings.
How to disable. In Claude Code remove the `calkit` plugin through `/plugin`. For other agents remove the skills installed by `calkit update agent-skills`. Project files and `calkit.yaml` stay.
FAQ
How do I install the Calkit skills in Claude Code?
Run `/plugin marketplace add calkit/calkit`, then `/plugin install calkit@calkit`. Commands such as `/calkit:create-pipeline` become available.
How do I add the skills to Codex or OpenCode?
Run `calkit update agent-skills`. The command installs the same skills into those agents' directories.
Why may the agent not save figures itself?
The plugin rule: the agent writes code, the pipeline creates outputs. A figure saved by hand has no record of environment and inputs, and `calkit status` cannot tell when it goes stale.
What does `check-reproducibility` check?
The skill runs `calkit check repro --json`: whether a pipeline and remote exist, which stages run outside an environment, which artifacts lack provenance and which manuscript numbers have no backing.
Do I need to install Calkit itself?
Yes. The plugin describes how to work with the CLI and does not replace it. Install with `uv tool install calkit-python` or the installer script from the README.
Discussion
No questions or reviews yet.
Related
PluginsProbabl skills for machine learning experiments: a scikit-learn and skrub pipeline, evaluation, audit and a link to Skore Hub
26 skills and four reviewer agents for science in Claude Code: figure and statistics review, literature search, work handoff.
Skills and a pipeline for economics and finance research: from a topic to a verifiable LaTeX paper draft.
Around 200 skills for Claude Code on bioinformatics, chemistry and biostatistics, from RNA-seq analysis to molecular docking
16 research methodology skills for agents: from a testable question and pre-registered hypotheses to reproducible analysis
24 skills for numerical methods and materials science: the agent checks stability, meshes, convergence and cluster runs.
README in short
The README presents Calkit as a one-command tool for research projects: literature, data, analysis and writing live in one repository. A separate section covers agent skills, and the project description says the pipeline lets you audit an AI agent's work through transparent, repeatable steps.