SkillHub Publishing Assistant

What it addresses

Publishing a local skill to the SkillHub community source often fails on engineering details rather than the API call itself: Git Bash paths like /c/Users/... are treated as relative paths, runtime and temporary files are packaged, SKILL.md frontmatter is incomplete, republishing the same slug and version returns 409, and large directories can exceed upload limits. SkillHub Publishing Assistant wraps these steps in a Python workflow that makes release updates easier to review and repeat.

How it works

• No separate CLI dependency: the script calls the SkillHub community publish API with Bearer Token authentication; it can reuse ~/.skillhub/credentials.json or accept SKILLHUB_TOKEN.
• Pre-release checks: run dry-run first to validate metadata and packaging, then publish; use --bump patch/minor to avoid overwriting the same version.
• Automatic exclusions: it excludes node_modules, venv, __pycache__, .env, logs, locks, sessions, and similar runtime files; .skillhubignore can declare skill-specific exclusions for local hook or daemon artifacts.
• Size checks: after packaging, it scans top-level directories and warns about unexcluded directories larger than 50MB, reducing the chance of hitting SkillHub upload limits.

Boundaries and caveats

This assistant is scoped to publishing or updating skills to the SkillHub community source; it does not delete or unpublish them. If platform deletion is needed, handle it through the SkillHub backend. After publishing, normal installs should pick up the new version; only if install or search still resolves the old version should the backend be used to set it as current. First-time publishing requires registration, identity verification, and an API Token.