Contributing

Contributions are welcome! This project is a Claude skill library for Higgsfield AI prompt generation.

How to contribute

Open an issue first for major changes so we can discuss the approach.
Fork the repo and create a feature branch.
Make your changes.
Test by running your prompts through Claude with the Higgsfield skill loaded — there's no traditional test suite, so verification means confirming the skill produces correct, high-quality Higgsfield prompts.
Run python3 scripts/validate.py before opening a PR. One-line environment bootstrap: python3 -m venv .venv && source .venv/bin/activate && pip install -r requirements.txt. Most tooling is pure stdlib, but the PDF user-guide generator (scripts/generate_user_guide.py) needs fpdf2; without it, scripts/validate.py reports the PDF dry-run as SKIP and still exits 0 — release builds use python3 scripts/validate.py --strict, which treats that SKIP as a failure.
Open a pull request with a clear description of what changed and why.

Keep sub-skill files focused on a single concern.
Follow the existing MCSLA formula structure for prompt-related changes.
Update CHANGELOG.md if your change is user-facing.
Model information must stay accurate to current Higgsfield platform capabilities.
Prompt examples should be production-quality and follow MCSLA structure.
Cross-references between skill files must resolve to real paths (python3 scripts/validate.py catches these).

By contributing, you agree that your contributions will be licensed under the MIT License.