feat(storefront): v5.0.0 — model_engineering category + medsci-modeling plugin + end-to-end repositioning

[Yoojin-nam]

v5.0.0 — storefront major for the medical-AI model-engineering lane

A material distribution change, not a label bump. The model-engineering lane was built additively across v4.x (Phases 1–4) plus Phase 5 (build-lane breadth, now in this release); v5.0.0 gives it a dedicated storefront home and widens the repo identity to cover it.

Storefront carve

• New category model_engineering ("Model Engineering & Validation") and new marketplace plugin medsci-modeling, carved out of "Data & Study Design" (medsci-data).
• The 6 lane skills — architecture-zoo, model-scaffold, model-validation, model-card, model-evaluation, mllm-eval — now group under their own catalog filter and installable plugin (/plugin now lists nine category plugins).
• Both catalog generators updated with their self-tests: gen_skills_catalog_json.py (category mapping + order) → 9 categories; gen_marketplace_json.py (plugin name + description) → 9 plugins. Generator self-tests assert plugin-per-category coverage dynamically (6/6 + 7/7 pass).

Repositioning

• README masthead + ROADMAP rewritten to the end-to-end identity: an end-to-end research tool to design → scaffold → validate → publish — for the clinical manuscript and the medical-AI model behind it.
• "Clinical AI model research engineering is in scope" is now explicit, while "not a general AI-scientist platform" (nor a diagnostic tool / autonomous author) is kept. The lane integrates MONAI / nnU-Net and never reimplements them or runs anything autonomously.

Release bookkeeping

• Version 4.11.0 → 5.0.0: CITATION.cff, package.json, metadata/distribution_manifest.json (all three agree; check_version_consistency green).
• CHANGELOG.md cut [5.0.0] - 2026-06-28 (Phase 5 content moved under it).
• .zenodo.json release record refreshed (stale "32 EQUATOR guidelines" → 38 + model-lane clause + discovery keywords).
• Counts unchanged: 51 skills / 41 detectors / 38 reporting guidelines. CI stays torch-free.

Gates (all green locally, CI-mirror)

validate_skills.sh (PII clean) · all gen_* --check (catalog/marketplace/detectors/skill_docs/distribution_manifest) · validate_catalog_consistency · check_version_consistency · check_domain_probe_sync --strict (17 modules) · check_frontmatter_schema · validate_routing_assets --strict · check_locale_inventory · check_npm_package_contents · catalog/marketplace/detectors/hero self-tests (6/7/7/28).

Deferred (separate, outward-facing — needs explicit auth)

• Hero mirror for a lane skill: creating a public Aperivue/<skill> repo is a publish action; not bundled here to keep this PR strictly in-repo and non-publishing.
• aperivue-web homepage lane section (separate repo PR).
• Release publication (tag + GitHub Release + npm + Zenodo) held for explicit per-release authorization.

🤖 Generated with Claude Code