Goal: when upstream changes (WordPress core releases, Gutenberg releases, docs updates), the repo should regenerate indexes and (eventually) open PRs that update affected skills/references.
- Indexes and matrices
- WordPress core version list (latest stable + recent).
- Gutenberg releases list (latest stable + recent).
- WordPress ↔ Gutenberg mapping table (derived from canonical docs where available).
- Routing metadata refresh
- Update
shared/references/*.jsonfiles only.
- Update
This keeps automation deterministic and reviewable before it starts rewriting skill prose.
- “Reference chunk regeneration” from upstream docs into
skills/*/references/*.md. - Task-shaped deltas (e.g. a new Gutenberg package, new block APIs, changes in theme.json schema).
- Semi-automated PRs that include:
- regenerated references
- updated checklists
- updated eval scenarios
shared/scripts/update-upstream-indices.mjs- Fetches upstream sources and rewrites JSON indexes in
shared/references/.
- Fetches upstream sources and rewrites JSON indexes in
- Schedule a workflow (daily/weekly).
- Run
shared/scripts/update-upstream-indices.mjs. - If
git diffis non-empty, open a PR with:- a summary of changes
- links to upstream release notes
- a checklist for human review (“does this impact blocks/themes/plugin workflows?”)
- Always run
node eval/harness/run.mjs. - Optional: use Agent Skills reference validator:
skills-ref validate skills/<skill-name>
The automation should prefer canonical sources and avoid scraping where possible.
- WordPress core releases and API endpoints (official WordPress APIs)
- Gutenberg releases (GitHub releases)
- WordPress developer docs (used for the WP↔Gutenberg mapping when no API exists)