Record GNPL re-scope: narration/projection; core-factor/gql superseded

[hyperpolymath]

What this is

The migration record and the live copy-paste command sheet still describe GNPL as a
rename of the "Glyph query language", extracted from two source paths. Both are now wrong,
and one of them is wrong in a way an agent could act on.

Companion to hyperpolymath/gnpl#6.

Measured

Claim in the record	Reality
gnpl — "migration pending"	Done. lithoglyph/gql-dt/ landed as squash-import 102c79e (PR #4)
gnpl = gql-dt/ + core-factor/gql/	Only the first moved. core-factor/gql/ is 5 Factor files, and gnpl contains zero .factor files

The second gap is now a settled decision, not an omission.

The re-scope

GNPL has been re-scoped as a narration/projection language — "what account does this
evidence support, told from whose stance, with what warrant" — built as a layer on top of
GQLdt and lowering to it:

GNPL  ──lowers to──▶  GQLdt  ──FFI──▶  Lithoglyph

core-factor/gql/ was the runtime/dynamic counterpart to GQLdt's compile-time proofs. A
typed narration layer lowering to GQLdt supersedes it, so it stays here as legacy
rather than being migrated. Rationale: docs/THEORY.adoc and docs/LITHOGLYPH.adoc in the
gnpl repo.

Changes

• lithoglyph/MOVED.md — splits extracted from not-extracted; records that the
  gql-dt/ copy here is no longer canonical; states the core-factor/gql/ supersession
  and why. (Its existing lith-monorepo caution is preserved.)
• REGISTRY.adoc — splits the combined row into GQLdt (nested in gnpl,
  canonical, lake build green) and GNPL (narration/projection), plus a note recording
  the re-scope and its consequences.
• RESITE-DATABASES-TO-OWN-REPOS.adoc — table row split; the gnpl decision marked
  DONE with the re-scope recorded as a follow-on decision; un-nest order corrected.
• RESITE-CODEX-HANDOFF.adoc — ⚠️ the load-bearing fix.

Why the handoff sheet matters most

That file is the agent command sheet — exact, copy-paste commands. Step 1d read:

rm -rf /tmp/gnpl && git clone --no-local /tmp/ngdb /tmp/gnpl
( cd /tmp/gnpl && git filter-repo \
    --path lithoglyph/gql-dt/ --path lithoglyph/core-factor/gql/ ... )
publish /tmp/gnpl hyperpolymath/gnpl "..."

Run today, that would git push a freshly-filtered history at
hyperpolymath/gnpl — which is already populated. Now marked DONE / DO NOT RUN with
the commands commented out and retained only as a record.

Two further corrections in the same file:

• Step 0's sanity ls -d … lithoglyph/gql-dt lithoglyph/core-factor/gql no longer requires
  paths that are either already extracted or deliberately staying.
• Step 1e's slimmed-parent filter no longer drops core-factor/gql/. It was superseded,
  not extracted, so it belongs to the slimmed lithoglyph — dropping it would have
  deleted live legacy code from the parent repo.

Also recorded

The gnpl extract was a squash, so granular history for that path survives only in the
_split_gnpl branch. Noted in the handoff sheet so it is not pruned.

Not touched

lithoglyph/CHANGELOG.md and lithoglyph/ROADMAP.adoc reference core-factor/gql/*.factor
— those are historical records of code that still exists here, and remain accurate.

🤖 Generated with Claude Code

---

Note on this PR's red checks — all pre-existing, none caused by this change

This PR edits only .adoc/.md files. The failing checks predate it:

• Check Workflow Staleness / Language / package anti-pattern policy — the standards
  reusable pins here (d7c22711e830) are 59 commits / 24 days behind standards HEAD,
  outside the recency window. Fix is scripts/propagate-workflow-pins.sh, separate from
  this change. The same job also flags the Scorecard/SARIF policy — filed as a template
  issue upstream, scorecard.yml template violates the estate's own Scorecard/SARIF policy as distributed standards#520.
• Trusted-base reduction policy — 20 escape hatches, undocumented in PROOF-NEEDS.md,
  and every one of them is in lithoglyph/gql-dt/src/GqlDt/ — i.e. the copy this PR
  records as no longer canonical.

That last point is worth stating plainly, because it is this PR's thesis showing up as a
CI failure: lithoglyph/gql-dt/ is now duplicated proof debt. The same axioms
(TypeSafe.lean:194, TypeSafeQueries.lean:44/73/99, …) are documented properly in
gnpl's new docs/proof-debt.md, and undocumented here.

Suggested follow-up (deliberately not done in this PR): per the RESITE procedure's own
"Replace, then thin" step — "only after the destination repo is verified, remove the
directory here" — lithoglyph/gql-dt/ should now be deleted from this monorepo. The
destination is verified (gnpl builds, CI green). Deleting live code is a bigger call than
a documentation correction, so it is flagged rather than taken.

Upstream issues filed during this pass: hyperpolymath/standards#516 (Hypatia
ImageOS=ubuntu24), #517 (check-trusted-base.sh Lean false positives), #520
(scorecard.yml template).