Version. 0.24
Date. 2026-08-04
Charter. standing-notes/dunin7-standing-authorization-charter-v0_1.
Author. Claude Code (execution session). Operator: Marvin Percival.
Supersedes. v0.23 at record b4e4a4d. All earlier versions stand as siblings; v0.3 remains deliberately absent.
Changes from v0.23. CR-2026-166 executed in full — Steps 0–6, Checkpoint A, Checkpoint B, in both repositories — and CLOSED. B-33, B-34, B-42 close. One new item, B-51, opens.
loomworks-engine main 7bdbf8b (was 3b10ff1), tag stele-mount-test-fix-v0_1, push-verified. Engine failure set now empty: 0 failed, 3522 passed, 68 skipped (was 1 failed).
loomworks main 7a9adc9 (was cb8b940), tag clean-report-surface-v0_1, push-verified. npm run lint: 0 errors, 0 warnings (was 33/23). npm run test: 696/696 pass, 0 unhandled rejections, exit 0 (was 12 rejections, exit 1).
Two repositories, separate branches, separate commit series, separate merges and tags throughout — never one commit spanning both.
B-33 (a test regression, not a defect). Stele's step-up auth work (TS-14) added a require_fresh_session slot to the passkey-enrollment route 17 days after test_stele_router_mount.py's fixture was last touched; the fixture never learned of it. Fixed by overriding the new slot too, confined to that one test file — no engine src/ change.
B-34 (a real gap, closed correctly). usePlatformLevel's effect called its own resolver fire-and-forget. Fixed by catching the rejection — but deliberately without setting resolved: true on failure, since every caller treats resolved: true, level: null as a confirmed absence of standing, not an unknown. A new test (usePlatformLevel.test.tsx) pins this with a load-bearing assertion: a network failure must leave resolved: false.
B-42 (verified site-by-site, not sampled). All 19 react-hooks/set-state-in-effect sites were individually checked against their own effect's dependency array — not just the two the Step 0 grounding sampled. None is a cascade; each got an inline exception naming its specific pattern.
One finding turned out larger than Step 0 read it, and was handled by deferring rather than forcing a fix or hiding it. react-hooks/static-components (Home.tsx) is a real defect — a component defined inside another component's render body, remounted with state loss on every parent re-render — but the correct fix touches four components and roughly six threaded closure values on a live dashboard screen, well past a "routine sweep." Rather than either rushing that refactor into this CR or excepting it with language implying the rule doesn't apply, the exception comments state the hazard plainly and defer the fix to a new item, B-51, named where a reader will find it.
Every exception this CR added is accounted for, in two honest categories, per the implementation notes: 29 verified individually as not applying (19 set-state-in-effect + 9 refs + 1 globals), 2 where the rule applies and the fix is named and deferred (static-components → B-51). Zero exists without one or the other.
Out of scope, confirmed untouched at every step: the engine's ruff backlog (3606 errors / 39 rules) and strict mypy backlog (601 errors), and the surface's direct-tsc --noEmit gap (4 errors). Neither repository "reports clean" in the fullest sense — this CR closed three of the six gates the Step 0 inspection found, exactly as its own §1 said it would and no further.
The Home.tsx nested-component defect (react-hooks/static-components), named above. Needs its own scoped change request: hoist LensBody, EngagementsZones, NeedsView, RecentView to module scope, threading zones, onOpen, needsList, recentList, addressById, and lens/loading/error as explicit props.
R-2 held through both checkpoints, in both repositories independently. Auto-proceed Steps 0–6 ran with per-step commits and per-step verification in each repo; Checkpoint A reported and proceeded on a clean report (no halt condition fired); Checkpoint B merged --no-ff, tagged, and pushed each repository separately under R-2's existing authorization.
C-1 held at both merges. Both repos' origin/main re-fetched and re-verified unchanged immediately before merging; clean index confirmed in each; --no-ff used in both; both feature branches deleted only after their merge landed and pushed.
A methodology error was caught and corrected, not reported as a false anomaly. The first attempt at the engine's Step 0 baseline accidentally ran against playground_test directly (40 apparent failures, from playground_test's schema predating a recent migration) — recognized as this session's own error rather than a genuine regression, and redone correctly on a throwaway per the CR's own kickoff fence. playground_test itself was never mutated.
Discipline clause held throughout. Every step's before/after numbers are quoted in the implementation notes. The static-components finding's true scope was reported honestly when discovered mid-session, rather than either silently downgraded to an exception or silently expanded into an unplanned refactor — the choice was surfaced and the resulting exception language says the true thing.
R-1 for this filing, on a clean pre-flight.
DUNIN7 — Done In Seven LLC — Miami, Florida DUNIN7 — status brief — v0.24 — 2026-08-04 Zero errors and zero warnings, readable because every exception behind them says which kind it is.