[JakeGinnivan]

Don't squash-merge this. Use "Create a merge commit" or "Rebase and merge". A squash rewrites the git-subtree-split trailer the next /core-pull needs to find where this one left off.

Pulls ddd-core through dddwa/ddd-core#30, which carried this fork's #204–#215 upstream, plus the fixes core needed to run them against its stub.

Summary

d7c21a6  Squashed 'core/' changes from 89be5d1..8a6c80a   (git-subtree-split: 8a6c80a)
70b2f5f  chore(core): pull ddd-core through #30           (merge + conflict resolutions)
139f875  chore(skills): link core's skills instead of copying them up

What arrives in core/. Mostly this fork's own changes coming back, plus what core added on top:

• a synthetic fixture conference year (E2E_FIXTURE_CONFERENCE, lib/e2e-fixture-conference.server.ts) that feedback-links and temporal-browser specs now opt into, so conference-day tests stop depending on a real year's dates. Years are read through getConferences(context) rather than conferenceManifest.conferences.
• tests that no longer assume the conference is in Perth or has a sponsor portal.
• MDX pre blocks take tabIndex={0} (axe scrollable-region-focusable).
• setup-* targets run with parallelism: false (the ENOTEMPTY race).
• /core-pull and /new-conference skill updates: Temporal migration notes, porting root-config changes, and linking skills.

Skills are symlinks now, not copies.

 .claude/skills/
-├── core-pull/SKILL.md        # copied up by scripts/sync-core-skills.mjs on prepare
-├── core-push/SKILL.md
-└── new-conference/SKILL.md
+├── core-pull -> ../../core/.claude/skills/core-pull
+├── core-push -> ../../core/.claude/skills/core-push
+└── new-conference -> ../../core/.claude/skills/new-conference
-scripts/sync-core-skills.mjs
 package.json                 # - "prepare", - "sync-skills"

ddd-core#30 makes forks link their skills, so a pull updates the skills in effect with no step to remember. That supersedes sync-core-skills.mjs, and its pnpm sync-skills step in core-pull was the one place this fork's core/ differed from upstream. core/ now matches ddd-core exactly. The fork's copies had no local edits (checked against core's pre-pull versions), so nothing was lost.

Conflicts (97), all expected. The last sync point predated #26, so everything this fork had upstreamed since came back as add/add:

• 24 code files → upstream. Each was the fork's version plus core-only fixes.
• 72 baselines → the fork's. Core's are rendered against the stub.
• core/nx.json → stays deleted.
• e2e/routes.ts auto-merged with HEADER_DATE_STATES declared twice. Removed the copy; the file is unchanged from main.

Evidence

Check	Result
nx build website	passes
typecheck, lint, test	pass, 945 unit tests
pnpm vr against this fork's baselines	108 passed, 36 skipped. Nothing in the pull changes how the site renders
nx e2e website	110 passed, 1 failed: my-agenda "picking the long talk displaces both short talks" timed out in dark mode, then passed 72/72 on --repeat-each 4. Flaky, not this pull
git-subtree-split trailer	present, 8a6c80a = ddd-core#30's merge

Merge Danger

Door: two-way

No new migrations: 0031_sponsor_feedback.sql has been in this fork since #204.

Blast Radius: small

• Windows checkouts. Git on Windows checks symlinks out as plain text files unless core.symlinks is on, and the skills then fail to load. The copy script didn't have that problem. Reverting 139f875 restores it.
• Conference-day e2e now runs against the fixture year, not 2026's config.

🤖 Generated with Claude Code