Read AGENTS.md for the full contributor guide (tech stack, boundaries, and commands).
- Git: use Conventional Commits in the form
type(scope): description, with types such asfeat,fix,docs,style,refactor,test, andchore(see AGENTS.md). - Pre-commit (recommended): this repo includes pre-commit in
.pre-commit-config.yaml(YAML checks and a commit-msg hook aligned with the rules above). Install once per clone:pip install pre-commitorbrew install pre-commit, thenpre-commit install --install-hooks. Details and manual checks are in AGENTS.md under Pre-commit hooks.
The unified skills-generator module holds all XML sources and Java code used to build agent skills under skills/.
- System prompt XML files use the PML Schema (pml.xsd). They are transformed with CursorRulesGenerator.java and system-prompts.xsl when producing reference content for skills.
- Skill summaries and inventory (
skills/,skill-inventory.xml) driveSKILL.mdgeneration.
If you have the idea to contribute, review the whole process in detail:
./mvnw clean verify -pl skills-generator # Pass tests
./mvnw clean install -pl skills-generator # Generate skills into skills/
npx skill-check skills # Validate generated skillsKeep skill-inventory.xml aligned with skills/ and system-prompts/ when adding or changing skills.
When you feel confident with the process, fork the repository and try to create new XML documents. Models will help you because an XML file is more rigid than natural language and it has a common vocabulary to create prompts.
When you feel confident with the solution, send a PR.