[malibio]

create_schema now refuses a name whose derived id is a valid YYYY-MM-DD date (e.g. "2026-10-06", "2026_10_06", "2026 10 06"), reusing the date node's id predicate. The check runs beside the core-type check, before any write. Digit-bearing names ("Invoice 2026", "Q4-2026") remain accepted.

Closes #3674

🤖 Generated with Claude Code

[malibio]

Code Review Summary

Small, well-scoped change (3 files, +40/-1) that does exactly what #3674 asks. The check sits beside the core-type check in create_schema, reuses the existing is_date_node_id predicate (widened to pub(crate)) rather than a new regex, and runs before any write. Net positive; no blockers.

Acceptance criteria

• Rejects date-shaped derived ids incl. "2026_10_06" / "2026 10 06" with an error saying the name cannot be a date: met (message is actionable, suggests "Log ").
• Digit-bearing names ("Invoice 2026", "Q4-2026") still accepted: met, covered by test.
• Runs before any node is written: met; placed right after normalize_schema_id/core-type check, before relationship validation and node creation. Test asserts no schema node exists afterward.
• Refusal and acceptance tests; refusal test fails without the check: met. Without the check fields: [] creates the schema, so expect_err panics for each of the three names.
• quality:fix: not verifiable here; confirm before merge.

Findings

Critical Issues

None.

Suggested Improvements

None blocking.

Nitpicks

• Nit: packages/core/tests/it/kebab_schema_id_test.rs: the "nothing left behind" assertion only checks 2026-10-06; since all three names derive the same id, that is sufficient, but a short comment saying so would help.
• Nit: the predicate also validates calendar semantics, so "2026-13-45" is accepted. That matches what a date node id is, so it is correct; no change needed.
• CLAUDE.md compliance: no literal node_type comparison, no lifecycle reads, no compat code, no wire type or schema change, so the node-type sequence and node-types.md do not apply. No new docs required (issue says none expected).

Recommendation

APPROVE. Run /address-review next per project workflow (expected: nothing to address).

[malibio]

Automated pragmatic-code-review: APPROVE (posted as a comment — GitHub blocks self-approval).

[malibio]

Addressed the one nit (comment in the test). Skipped none else; reviewer noted the calendar-validity behaviour of the predicate needs no change. Comment-only change, no re-review needed.