[umbrella] Estate audit & optimization — make the enforcement real (Waves 0–6)

[hyperpolymath]

Program tracker for the estate-wide audit-and-uplift anchored in standards. Ordered by estate doctrine (RSR-PHILOSOPHY.adoc): holes before goals, solutions at source, always fail loudly.

The diagnosis (7-agent recon): the estate's declarations were rich but enforcement was largely theatre — validators passed vacuously, mandatory checks were declared-but-never-run, and there was no post-action check on an agent's claims. Each wave converts a layer of theatre into a check that genuinely can fail.

Landed

• Wave 0 — kill the false green (Estate audit & optimization — Wave 0: kill the false green (holes before goals) #453): check-6scm.sh no longer passes vacuously; the boj-build K9 stub became a real structural check; rsr-audit.sh arg bug fixed (closes rsr-audit.sh: stale v1.0.0 canon + positional-arg bug → false-negative scores #387); the || true masking a broken RSR audit removed; audit-contractiles.sh de-hardcoded. Every fix ships a red-team fixture. Also fixed the root cause of estate-wide PR-red: main itself carried un-regenerated registry drift (relates REGISTRY.a2ml drift recurs and blocks every PR — needs a regen step or pre-commit hook #381/CI: governance + hypatia-scan red on main — 104 Hypatia findings vs empty baseline #399).
• Wave 1 — make automation run (Estate audit & optimization — Wave 0: kill the false green (holes before goals) #453): the Mustfile's 16 invariants now execute in CI (run-mustfile.sh); hooks/install.sh + just hooks-install; registry-drift CI now self-explains the fix; hypatia-scan honestly labelled ADVISORY (relates Proposal: an honest, adoptable readiness/insight badge framework — grade our own work, and challenge the evaluators #446).
• Wave 3 — compliance scorecards + dashboard (Estate audit — Wave 3: MUST/SHOULD/COULD compliance scorecards + generated dashboard #454): COMPLIANCE-DASHBOARD.md generated from 28 per-spec scorecards keyed to the registry, --check-gated. Honest baseline: 41/138 MUSTs passing, 66% systems coverage. Aspirational Grade-A gates never count as pass (operationalises Proposal: an honest, adoptable readiness/insight badge framework — grade our own work, and challenge the evaluators #446).
• Wave 4 — DYADT post-action claim verification (Estate audit — Wave 4: DYADT, post-action agent-claim verification (rein in LLMs) #457) + hardening (Harden the DYADT claim verifier against adversarial-review bypasses #458): the missing Tier-4 "agent claimed X → mechanically confirm/refute X". Reference verifier + conformance vectors + dogfood; hardened fail-safe against a 3-lens adversarial review.
• Wave 5 — per-language testing standard v2.0.0 (Estate audit — Waves 5 + 6: per-language testing standard, canonical-names guard, DYADT residual fix #459): RFC-2119 R1–R9 mapped to the CRG taxonomy; guide template; AffineScript guide (honest gaps); check-language-guide.sh gate.

Remaining charters (hand-off to other sessions / bots)

• DYADT production verifier — implement the continuous, network-capable verifier in hyperpolymath/did-you-actually-do-that against the conformance vectors in did-you-actually-do-that/spec/conformance/. Acceptance = passes that suite. (child issue)
• Per-language guides — Zig, Elixir+Gleam, Idris2/Agda from the template; refresh the Julia guide to R1–R9; migrate the AffineScript guide SSOT to hyperpolymath/affinescript. (child issue)
• Promote advisory scanners to blocking — flip affinescript-verify.yml and hypatia-scan critical-issue steps to gating against a committed baseline (relates CI: governance + hypatia-scan red on main — 104 Hypatia findings vs empty baseline #399, Hypatia scanner build-cache is stale-by-design — merged scanner fixes never activate in CI #437, Proposal: an honest, adoptable readiness/insight badge framework — grade our own work, and challenge the evaluators #446).
• Mustfile enforcement as a PR gate — run-mustfile.sh runs on push-to-main today; also gate PRs.
• 6a2 → descriptiles — complete the file migration + adopt a reintroduction guard estate-wide (rsr-template-repo); relates the CANONICAL-NAMES mandate.
• RSR v2.0 machine-half — port spec.scm/* (Guile, references deprecated STATE.scm) to .machine_readable/, reconciled with TIERS.adoc.
• Estate scorecard rollout — per-repo scorecards + an estate-wide grade dashboard.
• Release hygiene — first tagged release; rand<0.9.3 advisory expiry; audits/licence-flags-2026-07.adoc (flag-only).

Cross-links to existing issues

Relates: #381 (registry drift), #399 (CI-red on main), #426 (estate CI-red master cause), #437 (hypatia stale cache), #446 (honest badge framework), #451 (reusable-workflow permissions), #387 (rsr-audit bugs, fixed), #124 (proof-debt epic), #348 (gitbot fan-out).

Licence/SPDX is flag-only throughout every wave — no header edits.