Version. 0.21
Date. 2026-08-04
Charter. standing-notes/dunin7-standing-authorization-charter-v0_1.
Author. Claude Code (execution session). Operator: Marvin Percival.
Supersedes. v0.20 at record 75e759e. All earlier versions stand as siblings; v0.3 remains deliberately absent.
Changes from v0.20. CR-2026-165 CLOSED. Operator authorized the Step 3 correction in-scope (no v0.2 drafting session). Acceptance run re-run clean from a fresh throwaway on the corrected README. Merged --no-ff to loomworks-engine main, tagged, pushed. B-41 CLOSED.
loomworks-engine main 5c24d65 (was ffc29af), tag stand-up-sequence-v0_1, both push-verified (git ls-remote origin main / refs/tags/stand-up-sequence-v0_1 match local exactly).
Four commits merged: 07dd8e3 (the bootstrap script), 296f7f8 (README database section), 20a6953 (the Operator-authorized step-4 correction), 43adc6a (implementation notes at docs/phase-crs/cr-2026-165-implementation-notes-v0_1.md).
The Operator authorized the fix directly, in scope, per the queued decision at v0.20 — the cd-based form, not the untested -c alternative, matching the recommended default exactly. Not routed through a drafting-session CR v0.2, per explicit instruction.
The README's step 4 trap now states the mechanism, not just the instruction: uv run --project swaps the Python environment but does not relocate Alembic's own working-directory-based config discovery — named explicitly so a future tidy-up back to --project can't reintroduce the failure silently.
The acceptance run (§6) was re-run from the top, not resumed — a fresh throwaway (b41cr_acceptance2), all four steps, reading only the corrected README. All four steps completed clean; all three verification tests passed (test-suite table-count precondition, live app boot with /healthz reporting the administrative engagement present, both alembic heads matching their repositories' declared heads). Dropped and confirmed absent afterward.
Suite held at baseline across every run this CR triggered — Step 0, post Steps 1–2, and post the correction commit all returned 1 failed, 3467 passed, 68 skipped, the same long-standing test_stele_router_mount.py failure (B-33), unrelated.
A fresh Postgres instance can now be taken to a working database by following what loomworks-engine's own README says — the gap the scoping note (loomworks-b41-scoping-note-v0_2) named as the actual deliverable is filled, and it was proven filled by having an execution session follow only the page, not by asserting that it would work.
The gate did what it was built to do. The first acceptance run caught a defect in the CR's own deliverable — a command translated from the verification's proven form into an untested one, which failed exactly where an untested command was flagged as the risk. The halt-and-queue discipline (C-3) meant the fix was reviewed and authorized rather than silently patched past, and the second acceptance run is what actually closes the gate — not the first attempt, not the fix being plausible.
R-2 held through both checkpoints. Checkpoint A's halt condition (acceptance-run failure) fired correctly and halted rather than being absorbed. The Operator's authorization was the exact form required to resume — an explicit, scoped instruction, not a standing default. Checkpoint B — merge, tag, push — ran under its existing R-2 authorization once the gate passed clean.
C-1 held at the merge. origin/main re-fetched and re-verified unchanged (ffc29af, no concurrent-session drift) immediately before merging; clean index confirmed; --no-ff used, provenance preserved; the feature branch deleted locally only after the merge landed and pushed.
Discipline clause held throughout. Every command in this session's two acceptance runs is quoted with its output in docs/phase-crs/cr-2026-165-implementation-notes-v0_1.md. The second acceptance run was explicitly re-run from the top rather than assumed to inherit the first run's passing steps.
R-1 for this filing, on a clean pre-flight.
DUNIN7 — Done In Seven LLC — Miami, Florida DUNIN7 — status brief — v0.21 — 2026-08-04 The gate caught its own drafting error, and the second run — not the fix being plausible — is what closes it.