T20 schema approval is still a gate (blocking)

[thedancingdeveloper]

Issue #58 (T20) remains open. The MariaDB baseline and dependent T21-T26 work exist in the rescue branch, but the handoff explicitly requires human acceptance of the target schema.

The harness queue is paused rather than treating the generated baseline as approved. Nothing downstream of T20 should be accepted until this is decided.

This is a human decision, not work an agent can complete.

---

Raised from the handoff register: docs/2026-08-03-handoff-issues.md (NGMS repo).

[thedancingdeveloper]

Not resolvable by an agent — this is the human gate, working as designed.

T20 (#58) is open, and the register is right that nothing downstream should be accepted until the target schema is. The harness now enforces the shape of this rather than relying on discipline: agent-harness#80 added a preflight that refuses to start a project that cannot reach its definition of done, and the NGMS queue stays paused with 49 pending / 0 running.

What you need to decide: accept the generated MariaDB baseline, or revise it. Everything in T21–T26 is written against it, so the answer changes whether that work is kept or redone.

Leaving open.

[thedancingdeveloper]

🤖 Vogt validation pass — 2026-08-20 (automated review vs dd26a0aa)

Still a live gate, needs maintainer judgment. T20 (#58) is OPEN, so the schema-approval gate genuinely still holds. origin/main still ships crates/stackarr-postgres and has no stackarr-mariadb crate; the generated MariaDB baseline lives only on the unmerged feat/mariadb-baseline branch (main is 9 commits ahead of it). The pending decision — accept the generated MariaDB baseline or revise it — cannot be settled from code and gates whether T21–T26 are kept or redone. Leaving open for review.