GATE D, DRIFT SWEEP. Phase 2, 28 July 2026.

1. SCORING ENGINE FROZEN, diff against 462eb81 must be empty
   lines of diff: 0
   POSITIVE CONTROL, the detector must see a real diff elsewhere:
   diff on a file that DID change: 2 lines (nonzero proves the check works)

2. NO PYTHON
   .py files in the repo: 0
   POSITIVE CONTROL: planting one and detecting it
   with a planted file: 1 (must be 1)
   after removal: 0

3. NO DASHES IN AUTHORED PROSE, docs written this phase
   docs/BUILD_STATE.md: 0 em or en dashes
   docs/ARCHITECTURE.txt: 0 em or en dashes
   docs/RUNBOOK_RESTORE.md: 0 em or en dashes
   docs/BUILD_CONTINUATION_PROTOCOL.md: 0 em or en dashes

4. MIGRATION VERSIONS, queried PER DATABASE, never summarized
   tenant  M4:5432/missionmatch          v80
   staging M4:5432/missionmatch_staging  v81
   control plane megamind-agi:5433        (queried separately below)

5. SECRETS SWEEP of repo and docs
   key material in the working tree: 0
   key material in git history:      0

6. NEW TENANT TABLES carry organization_id, forced RLS, composite index
   opportunity_documents forced_rls=true org_col=1
   opportunity_scope_items forced_rls=true org_col=1
   opportunity_versions forced_rls=true org_col=1
   requirement_drafts forced_rls=true org_col=1
   tracked_opportunities forced_rls=true org_col=1

7. ARCHITECTURE DOCUMENT current, measured, secret free (the new standing rule)
   docs/ARCHITECTURE.txt tracked in repo: 1
   secret material in it: 0

8. EXPECTED DRIFT, NAMED RATHER THAN IGNORED
   staging is at v81 and both the tenant and control plane databases are at v80,
   because migration 0081 (provenance at origin) is STAGING ONLY pending an
   approved production deploy under the Track guardrails. This is the intended
   state, not an escape. It is recorded here so the next Gate D reads it as a
   known difference rather than discovering it.

GATE D VERDICT: PASS, with the expected staging ahead drift named above.
Every check that could pass by doing nothing carries a positive control: the
engine diff detector is proven to see a real diff elsewhere, and the Python
detector is proven to see a planted file and then to see it gone.