CANON STATUS — generated; do not hand-edit

Declared source: OFFERING_MAP v2.3 §B.6 (machine copy: contracts/mizoki_contracts/canon_status.py) · Generator: scripts/gen_canon_status.py · deterministic (no timestamps — the doc changes only when the declared table or the probed source changes).

Every IN BUILD row's flag posture is probed from SOURCE literals — the same fail-if-flipped discipline the writeback flags carry. Drift between a declared status and its probed truth is listed below and printed as a CI WARNING by the governance suite; it never silently passes.

Capability Declared status Gate note Code-truth probes
SRPVDAL state machine, Decision Control Plane, DEL scoring, autonomy ladder, audit ledger LIVE DEL platform floor 80.0; Stage-3 recommend-only fleet default — (no flag probe declared)
Connector gateway & canonical envelope LIVE one governed receiver per provider; consent fail-closed — (no flag probe declared)
Signal public surface, capability dossiers, Signal Factory demo (Gate 1) LIVE site ships only via human APPROVED dispatch + canon check — (no flag probe declared)
Causal proof core (meta-learners, refutation, triangulation) PARTIAL serving lanes per registry; first real registered holdout open — (no flag probe declared)
ORACLE intent cells 33–36; Growth Decision Graph intent edges PARTIAL observe-only / shadow; promotion gates Brier ≤ 0.20, AUC ≥ 0.72, ≥ 2 cycles, human decision — (no flag probe declared)
Intent Engine v2 modules I-01…I-05 IN BUILD shadow, flags off, retention bounds as code ok: src/cells/cell33/ingest_cell/config.py: 'False'
ok: src/cells/cell34/scoring_cell/config.py: 'False'
ok: src/cells/cell35/graph_cell/config.py: 'False'
ok: src/cells/cell35/graph_cell/config.py: 'False'
Net Yield lane (order economics, cohorts, writeback dry-run) IN BUILD writeback off until verified pilot (fail-if-flipped tested) ok: services/net-yield/writeback/init.py: 'false'
Measurement rails writeback PARTIAL MEASUREMENT_WRITEBACK stays OFF (ast-pinned literal) ok: services/measurement-rails/flags.py: 'False'
F3 supply-chain / inventory sync IN BUILD observe-only (ruling 2026-08-19-A); recommendations can never reach a platform adapter (hard invariant) ok: src/cells/cell37/market_cell/config.py: '"F3_INVENTORY_SYNC"'
F4 auto micro-geo calibration LIVE pilot armed 2026-08-24 (mycocoons; F4_CALIBRATION=true); every reservation L2-approval-gated; bounded autonomy only after 2 clean cycles (0 complete: first reservation awaits owner L2, OPEN_ITEMS P-1; actuator deliberately unregistered, P-2); unconfigured tenant vault = lane dark (not_configured); source default stays OFF (probe). Ruling 2026-09-13 (owner-delegated) supersedes the 2026-08-19-B IN BUILD label ok: src/shared/growth_control/f4_calibration/reservation.py: 'False'
F1 creative unbundling IN BUILD estimates provisional until pilot creative volume; generated-creative deployment retains human approval always ok: src/shared/creative_aesthetic/elements.py: 'False'
F2 multi-quarter LTV regimes IN BUILD findings gated on ≥ 2 observed quarters per cohort — the guard is code (data_insufficient), never an extrapolation ok: src/shared/growth_control/f2_ltv/dtr.py: '2'
F5 treasury-gated spend governance IN BUILD v1 config-declared floors, fail-closed (absent config = no constraint claimed); v2 on live treasury feed [IN BUILD-v2] — (no flag probe declared)
90-Day Pilot & packaging ladder ADOPTED commercial standard (ruling 2026-08-19-A); the pilot report produces the verified numbers that flip Preview labels — (no flag probe declared)

Drift register

Duplicate-topic register

A row with no probe is declared-only (LIVE/PARTIAL rows rest on the registry and dated live-verification records, not on a single flag literal). The registry outranks this file; this file outranks prose.

← All docsView source on GitHub →