A Codex skill project for helping AI agents write, review, parameterize, and maintain mumax3 .mx3 simulation code.
This repository contains a cleaned generic mumax3 skill. Project-specific research material has been removed from the public snapshot; the installed skill path is kept generic for broader reuse.
MumaxWriter.skill/
README.md
README.zh-CN.md
README.ja.md
examples/
prompts/
expected_outputs/
catalog/
manual_evaluation_checklist.md
skills/
mumax3/
SKILL.md
agents/
openai.yaml
assets/
templates/
minimal_relax.mx3.template
thin_film_with_region.mx3.template
local_ac_field.mx3.template
references/
language_subset.md
anti_patterns.md
api_index.md
api_patterns.md
api_full.md
examples.md
structures.md
parameter_sweep.md
output_and_reproducibility.md
compatibility_notes.md
scripts/
check_mx3_static.py
render_template.py
collect_case_status.py
docs/
release_notes_v0.1.0.md
third_party_sources.md
troubleshooting.md
tests/
validate_release.py
The mumax3 skill helps an AI agent:
- Write valid mumax3
.mx3scripts instead of full Go code. - Review scripts for common mumax3 syntax and API mistakes.
- Parameterize mumax3 scripts for batch simulations.
- Structure scripts around mesh, geometry, materials, initial states, fields, damping, output, and run stages.
- Decide when to use Python to generate scripts, run mumax3, organize outputs, fit results, and plot figures.
Copy the skill folder into your Codex skills directory:
Copy-Item -Recurse .\skills\mumax3 C:\Users\<you>\.codex\skills\mumax3Then invoke it explicitly with:
Use $mumax3 to write or review this mumax3 script.
Install the standalone skill from GitHub with:
npx codex-marketplace add msjs2333/MumaxWriter.skill/skills/mumax3 --skillThen invoke it explicitly with:
Use $mumax3 to write or review this mumax3 script.
Use $mumax3 to write a minimal mumax3 thin-film relaxation script with SI units, table output, and a final OVF save.
For parameter sweeps, ask the skill to keep Python responsible for case generation, metadata, logs, and postprocessing while keeping each simulation in native mumax3 .mx3 scripts.
Run local validation from the repository root:
python tests\validate_release.pyThe validator uses only the Python standard library. It compiles helper scripts, checks template parameter examples, runs conservative static checks on expected-output .mx3 files, and runs mumax3 -vet when mumax3 is available on PATH.
This is the v0.1.0 public snapshot. It includes generic mumax3 templates, expanded API pattern references, deterministic helper scripts, and a repository-level evaluation surface under examples/.
The examples and static checks are intended for skill validation and script hygiene. They do not prove physical correctness, numerical convergence, or publication readiness.
If you use this skill in research, teaching, or published workflows, please cite the repository using the metadata in CITATION.cff. GitHub should show this metadata through the repository's "Cite this repository" button.
This project is prepared for Zenodo software archiving. After Zenodo assigns a DOI for the v0.1.1 release, add the DOI badge and DOI link to this section.
Planned directions for later releases:
- Evaluate optional Python-side data handling workflows, including possible use of the Ubermag ecosystem and related analysis tools.
- Add mumax+ compatibility, with the long-term direction of making mumax+ the primary target while keeping mumax3 support.
- Provide assisted migration workflows from existing mumax3
.mx3scripts to mumax+, including API mapping notes, behavior differences, and validation checklists. - Expand postprocessing examples for OVF/table output, spectra, mode maps, texture tracking, and sweep summaries.
- Release notes v0.1.1: citation metadata update and Zenodo archiving notes.
- Release notes v0.1.0: initial public snapshot scope, compatibility, validation status, and known limitations.
- Example catalog: prompts, expected outputs, and coverage map.
- Third-party sources: external source and provenance notes.
- Troubleshooting: common local validation and mumax3 issues.
Licensed under the Apache License 2.0.