Contributing to MAGIC
MAGIC Agent Skills is open-source and community-driven. Here is how to contribute to any suite in the ecosystem.
Which Repo?
Data Agent Skills
Data science skills: loading, profiling, cleaning, transforming, analyzing, visualizing, reporting.
github.com/Votee-AI/magic-data-agent-skills →Linguistic Agent Skills
NLP skills: multilingual processing, corpus curation, linguistic analysis, low-resource tooling.
github.com/Votee-AI/magic-linguistic-agent-skills →Universal Process
Regardless of which suite you contribute to, the process is the same:
- Pick an issue — browse the Votee-AI GitHub organization and look for
good first issueorhelp wantedlabels - Fork and branch — fork the repo and create a branch from
main - Write the skill — follow the universal SKILL.md anatomy
- Run skill-judge — address all P0 and P1 issues before submitting. See evaluation framework
- Open a PR — include real usage examples and acceptance evidence in the PR description
Skill Quality Bar
All contributed skills must:
- Follow the universal SKILL.md structure with all required sections
- Have verifiable acceptance criteria (not vague descriptions)
- Include at least one real usage example (not synthetic)
- Pass skill-judge with no P0 issues
- Fit within the domain scope of the target suite
New Suites
Want to start a new domain suite? Open an issue in the relevant repo on the Votee-AI GitHub organization with the tag new-suite and describe the domain, proposed skill count, and your intended contribution level. The maintainers will help scope and bootstrap the new suite.