v0.1.0

Highlights

• Launches the fhir-format skill focused on safer FHIR reasoning and formatting support.
• Adds explicit claim scoping in responses:
  • base specification behavior
  • version differences
  • profile-dependent constraints
  • implementation guidance
• Reinforces source-aware guidance and clearer validation boundaries (structural validation vs full IG/profile conformance).

Documentation & Publishing

• Added marketplace-ready plugin metadata in .claude-plugin/.
• Added polished GitHub Pages landing site (built from docs/page).
• Added submission helper content in MARKETPLACE_SUBMISSION.md.
• Added contributor and legal docs (CONTRIBUTING.md, LICENSE.md).

Quality & Reproducibility

• Added reproducible relevance benchmark (benchmark/) with frozen reference artifacts.
• Added CLI grounding benchmark lane (benchmark/cli/) for command-based evaluation.
• Added benchmark result validator (benchmark/validate_results.py).
• Added metadata consistency validator (scripts/validate_metadata.py).

CI/CD

• Added publish-readiness CI checks for metadata, benchmarks, and docs build.
• Added GitHub Pages deployment workflow.
• Added release workflow to enforce tag/version alignment before publishing GitHub releases.

Notes

• Initial stable baseline release for marketplace submission and community validation.
• No breaking changes (first release).

Full Changelog: https://github.com/dnlbox/fhir-format-skill/commits/v0.1.0