Dispatch Queue

orchestrator reconciler plan · unstamped
Build queue (1L)
2
→ model-developmentline:1L · build
Remediate queue (2L)
43
across models23
→ model-remediationthen 2L re-review
Intake queue
1
in_development→ 2L intake

Orchestrator computes this each tick from the git-issue queue; a runner executes the workflows (operator now; gated lifecycle-runner later). Output-changing / Tier-1 stays human-gated.

Flow

in-flight & completed work, by line of defense · source: git issues (539 tracked)
1L — Builder
0
open / in-flight0
completed14
build & remediation (line:1L)
2L — Reviewer
81
git mirror (open)45
open RAT conditions25
open inventory findings56
supplemented from attested 2L artifacts
3L — Controls Audit
4
open / in-flight4
completed1
controls-audit oversight (line:3L)
Throughput
470
total open69
total completed470
unlabeled (blind spot)20
closed kept as the done record

Recent flow activity

last 15 issue events
when repo line title state
2026-06-30 16:57:05Z modelriskapp Portal repo needs a control-plane auto-merge lane (stop hand-merging modelriskapp PRs) closed
2026-06-30 15:04:02Z insightalm 3L quality-check 40-turn flake should auto-rerun (stop hand-reruns blocking clean producer PRs) open
2026-06-30 13:41:54Z insightalm 3L D056 lane: add scripts/vm_migration/** to control-plane-review paths (it's provisioning machinery) open
2026-06-30 12:41:44Z insightalm 3L Observe-loop repo-red: separate main-CI health from open-PR check failures (don't count stale/behind checks) open
2026-06-30 12:14:29Z modelriskapp Portal hardening: drift guards run only in sync-data, not on PRs — can't gate the control-plane lane open
2026-06-30 07:38:48Z insightalm 3L Observe-loop: portal anomalies open
2026-06-30 05:56:30Z modelriskapp Portal repo needs a control-plane auto-merge lane (stop hand-merging modelriskapp PRs) closed
2026-06-30 04:21:38Z insightalm 1L [TEST build-intake] M-9TEST: lifecycle-sweep → 1L-build path proof (close after verify) closed
2026-06-30 04:10:33Z insightalm 1L [TEST build-intake] M-9TEST: lifecycle-sweep → 1L-build path proof (close after verify) closed
2026-06-30 03:31:20Z FinView CI fragility: binding-manifest B-06 check drifts because platform-lib is pinned to @main (moving target) open
2026-06-29 20:04:49Z insightalm 3L sre-telemetry PRs can land BLOCKED — required structure-check never triggers closed
2026-06-29 19:55:46Z insightalm Heartbeat Layer-5 false-positive: vm-workflows flagged STALE while runs succeed closed
2026-06-29 19:45:07Z insightalm 2L [2L] LG-055-07 (M-055): closed
2026-06-29 19:45:06Z insightalm 2L [2L] LG-055-06 (M-055): closed
2026-06-29 19:37:27Z insightalm Post-RAT sync drift: M-422 (banking) + M-432 (energy, →active) pointers unapplied — fails structure-check open

By model

labeled flow, grouped by the model the work attaches to
modelopencompleted
M-002 2
M-003 1 4
M-004 1
M-005 3 2
M-007 1
M-008 1
M-040 1 0
M-041 5 3
M-051 1
M-052 1
M-053 1
M-054 6
M-055 3 7
M-070 1 0
M-072 2 0
M-090 2 0
M-110 1
M-132 2
M-133 2
M-134 1
M-135 1
M-136 1
M-200 1
M-420 4 0
M-421 3 1
M-422 4 0
M-423 2
M-430 5 0
M-431 4 1
M-432 1 0

2L open conditions

open COND-* on the latest non-approved RAT per model — the attested 2L source
model cond deadline summary
M-005 COND-001 2026-09-04 NARROWED to (c), HELD on documentation-integrity. (a) orchestrator + (b) reinsurance-aware contract/rebind cleared earlier. The COND-001(c) SUBSTANCE (top-3 sensitivity + a compos…
M-040 COND-003 2026-08-23 The M-040 validation evidence pack is structurally complete and real, but one §10.5 item-5 sub-requirement is honestly disclosed as PENDING: a backtest of the CECL allowance again…
M-041 COND-002 2026-09-04 Resolve the output-contract posture. EITHER (a) explicitly ADOPT the by-design no-MRM-contract posture (outputs consumed in-process; the reproducing gold snapshot is the determini…
M-041 COND-003 2026-09-04 Finalize the assumptions posture. EITHER register the A-NNN (credit-spread/OAS basis, discount basis, total-return measurement, participation) with calibrated values + sources and…
M-041 COND-004 2026-09-04 Bring the validation pack to adequacy: it already has 11/11 tests + a reproducing gold snapshot + the SR-11-7 archive (real). Add the disclosed-firm embedded-derivative reconcilia…
M-041 COND-005 2026-09-04 §8.3 enum extension DONE — asc_815 is a registered framework (line 608) and a Tier-1 trigger (line 407). NOT cleared: re-tag M-041.regulatory_frameworks to add asc_815 (it is the …
M-055 COND-055-04 2026-09-04 HELD (now at the Tier-1 bar). The build mechanics are validated (composition cross-check 0.00e+00, monotone/unimodal sensitivity, deterministic snapshot), but the headline consoli…
M-070 COND-003 2026-08-23 Tier-1 validation-evidence ADEQUACY gap — UNCHANGED from RAT-070-v1.0.1, independently re-verified as STILL OPEN at this re-review (2026-06-08). The assembled pack and the model c…
M-072 COND-001 2026-08-23 NARROWED, STILL OPEN (the remaining Tier-1 blocker). The validation-evidence pack provides adequate conceptual soundness, multi-config/OOS, sensitivity on the material assumptions…
M-090 COND-001 2026-09-04 STILL OPEN (honest-conditional). No backtest evidence reconciling the attributed source-of-earnings components to a firm's realized earnings movement. Requires actual-vs-expected …
M-420 C-420-01 INTELLIGENCE-USE-ONLY. M-420 is approved ONLY for cross-firm banking assumption recovery + valuation (decision-support intelligence). It is NOT approved for any booked, regulatory…
M-420 C-420-02 REAL ENGINES ARE THE PRECONDITION FOR BOOKED USE + TIER PROMOTION. Any booked/regulatory banking use requires the PLANNED (not built) engines — M-421 CECL allowance (ASC 326, Tier…
M-420 C-420-03 RE-RATIFY ON ENGINE REPLACEMENT. When the analytic forward() is replaced by a real banking engine (the Phase-2c upgrade path), M-420 (or its successor) re-enters ratification at t…
M-421 C-421-03
M-421 C-421-04 2026-09-28 SEGMENT-LEVEL USE STAYS DECISION-SUPPORT ONLY (firm-level + implied-Q approved in-scope). The segment-decomposition validation (C-421-03) was performed and is NEGATIVE for concent…
M-422 C-422-01 MERGE-STATE / GOVERNANCE-VS-BUILT MISMATCH. The bottom-up HC-R engine (the basis for the insightalm registry's bottom-up promotion narrative, PR #3506) is NOT on bankview origin/m…
M-422 C-422-02 BOTTOM-UP IS DATA-EMPTY. The "regulator-reported, confidence 1.0" HC-R cells are an illustrative hardcoded JPM grid; no BHCF file has been ingested (the FFIEC NIC channel is CAPTC…
M-422 C-422-03 TIER-1 VALIDATION DEPTH. The merged disclosed-anchor tier reconciles RWA to the disclosed total BY CONSTRUCTION (D solves the level -> 0% error); the genuine signal is only the CE…
M-430 C-430-01 INTELLIGENCE-USE-ONLY. M-430 is approved ONLY for cross-firm utility assumption recovery + valuation (decision-support intelligence). It is NOT approved for any booked, regulatory…
M-430 C-430-02 REAL ENGINES ARE THE PRECONDITION FOR BOOKED USE + TIER PROMOTION. Any booked/regulatory utility use requires the PLANNED (not built) engines — M-431 regulated rate-base / cost-of…
M-430 C-430-03 THE PROXY / FREE-θ CAVEATS MUST STAY DISCLOSED. The two scaffolded substitutions — (a) the invested-capital proxy rate base for firms lacking a utility_plant_net series (D, NEE) a…
M-430 C-430-04 RE-RATIFY ON ENGINE REPLACEMENT. When the analytic forward() is replaced by a real regulated-utility engine (the M-431/M-432 upgrade path), M-430 (or its successor) re-enters rati…
M-431 C-431-01 RECONCILE THE GOVERNANCE ARTIFACTS TO THE BUILT STATE. The engine energylib.L3.cost_of_service is built, tested, and honest — but the insightalm registry entry M-431 is still stat…
M-431 C-431-02 INTELLIGENCE-USE ONLY + KEEP THE RECOVERED/PROXY CAVEATS DISCLOSED. M-431 is approved ONLY as cross-firm cost-of-service intelligence (reconciliation + comparison). It is NOT a bo…
M-431 C-431-03 THE TWO MISSES MUST STAY DOCUMENTED AS SIGNAL, NOT SUPPRESSED. XEL (+17.2%) and NEE (−15.6%) do not reconcile within ±10%, and that is CORRECT and informative: NEE carries unregul…

Automation catalog

the machine, read dynamically — not hand-maintained

Workflows

5 lifecycle workflows (.claude/workflows)
workflowwhat it does
demand-arc Run the D042 demand-prescription arc end-to-end without a human picking the vertical: the demand-builder agent role (insightalm .claude/agents/demand-builder.md) proves a live vertical\
phases: Preflight → Select → Build-twin → Rank → Prove-firm-specific → Land → Self-merge
depth-arc Run the frontier-toll DEPTH arc end-to-end without a human picking the toll: the depth-builder agent role (insightalm .claude/agents/depth-builder.md) deepens an existing vertical at its next un-deepened load-bearing frontier toll. The depth-arc driver (tenk2win/scripts/run_depth_arc.py) reads ecosystem-mapper focus.json, picks the top toll above the 0.50 threshold with free public SEC facts (D062), and the role runs the playbook (fit-or-extend-θ → recover + discrimination check → flip the depth metric). Moves depth_proven by one toll; the spike PRs self-merge via the fleet auto-merge lane. Idempotent: re-running picks the next toll, skips depth-proven ones, honestly stops at the threshold or when remaining tolls are un-sourceable. The codified form of the arc hand-run 4x (EFX/MA/SCHW/FNMA).
phases: Preflight → Select → Fit-or-extend-θ → Recover → Flip-metric → Land → Self-merge
model-development Run the 1L model-development flow (Decision 053): the model-builder agent turns strategic demand (a model spec or an INV capability-gap finding) into a built model — engine + ENGINE_CONTRACT + legacy_metadata + card + bindings — then runs the gates and hands an intake-ready model to the 2L intake gate. Builds real code; never reviews or ratifies.
phases: Preflight → Scope → Design → Design-review → Build → Verify → Land → Signal → Synthesize
model-lifecycle Run the model lifecycle (Decision 052): the 2L reviewer agent reviews / ratifies / re-reviews / monitors models against the MRM standard, issues findings, drafts RAT verdicts — never edits code
phases: Preflight → Enumerate → Review → Land → Synthesize
model-remediation Run the 1L findings-remediation flow (Decision 053): for each conditionally-approved model, the model-builder agent reads the 2L COND-NNN gating conditions, classifies them (doc / validation-pack-assemblable / engine-build), remediates the assemblable ones with REAL evidence, reconciles registry doc flags, and reports which conditions are now ready for 2L re-review vs which stay honestly conditional. No fabrication.
phases: Preflight → Enumerate → Remediate → Verify → Land → Signal → Synthesize

Scheduled crons

19 scheduled workflows (.github/workflows)
name repo cadence detect / act
Binding Audit insightalm 30 4 * * * detects
Build Governance View insightalm 0 */6 * * * detects
Canonical Market-Data Refresh insightalm 30 6 * * * detects
D029 Reconciliation Report (Weekly) insightalm 30 6 * * 1 detects
Data Engineering Pipeline insightalm 0 */4 * * * detects
Demand Arc insightalm 30 9 * * 3 acts
Depth Arc insightalm 30 8 * * 2 acts
MRM 2L → Git Flow Mirror insightalm 20 7 * * 2 acts
MRM Closure-Loop Registry Sync insightalm 30 6 * * * acts
MRM Ratifications Audit insightalm 0 5 * * * detects
Model Lifecycle Monitor insightalm 30 7 * * 1 acts
Observe Loop insightalm 20 */6 * * * acts
Quality Engineering E2E insightalm 0 2 * * * detects
SRE Telemetry insightalm */30 * * * * detects
Security Runtime insightalm 0 6 * * * detects
Use Case Validation Flows insightalm 0 */6 * * * detects
auto-rebase-stale-prs (self-heal the stale structure-check red) insightalm 17 */4 * * * detects
lifecycle-sweep insightalm 20 1-21/4 * * * detects
remediation-sweep (17th agent) insightalm 40 */4 * * * detects

Event-triggered lanes

11 on-demand / PR-triggered workflows (.github/workflows)

Autonomous lanes that fire on an event rather than a clock — the 1L remediation runner (D055) and the control-plane review-and-merge lane (D056). These act without a human; output-changing / Tier-1 and reviewer-self edits stay gated.

name repo trigger detect / act
CI InsModel on PR, on push detects
auto-merge (clean PRs through the gates) InsModel on PR, manual acts
CI goes on PR, on push detects
CI insightalm on PR, on push detects
PR Quality Check insightalm on PR, on @mention detects
control-plane-review insightalm on PR acts
lifecycle-runner insightalm manual acts
merge-clean-prs (operator cockpit — REPORT-ONLY) insightalm manual detects
queue-drainer (operator cockpit) insightalm manual detects
remediation-lane insightalm manual acts
structure-check insightalm on PR, on push acts

Governance tooling

58 scripts (scripts/*.py gates)
scriptpurpose
aggregate_binding_manifests.py Aggregate per-repo binding manifests and audit against registries.
auto_rebase_stale_prs.py auto_rebase_stale_prs.py — the durable self-heal for the "stale structure-check
backfill_canonical.py One-shot CLI for backfilling canonical-market-data observations.
backfill_eris_swaps_from_legacy.py One-shot: preserve legacy Eris SOFR-OIS swap history in canonical (D029 #921 prep).
backtest_runner.py Back-testing framework for FinView pricers [E4].
briefing_aggregator.py Deterministic briefing snapshot — reads every tracking/agent_state/*.yaml
briefing_aggregator_selftest.py briefing_aggregator_selftest.py — proof for the VM-staleness classification (#3051).
build_governance_view.py build_governance_view.py — Render governance artifacts to HTML viewers.
canonical_market_data_check.py Canonical market-data layer inspector — emits findings to agent state.
changelog_tool.py changelog_tool.py — derive a Keep-a-Changelog [Unreleased] section from git history.
check_db_registry.py DB-registry guard (security control) — code may only open a registered .db.
check_firm_facts_single_source.py check_firm_facts_single_source.py — enforce Decision 047.
check_registry_integrity.py Registry-integrity guard — models / engines / assumptions may not drift or go untracked.
check_registry_reality.py check_registry_reality.py — registry status must not contradict on-disk reality.
check_role_boundary.py check_role_boundary.py — technically enforce separation-of-duties on a PR's
check_role_boundary_selftest.py check_role_boundary_selftest.py — proof harness for check_role_boundary.py.
check_status_guard.py check_status_guard.py — a 1L (or any producer) registry-delta may NOT promote a model
check_structure.py check_structure.py — enforce operations/structure_map.yaml against disk.
check_validation_pack_honesty.py check_validation_pack_honesty.py — fail-closed guard against validation packs
consistency_checker.py InsightALM Registry Consistency Checker — detects drift between modelling-
contract_validator.py InsightALM Contract Schema Validator — validates data against contract JSON Schemas
data_engineering_check.py Data engineering pipeline checks — runs on the VM via GHA self-hosted runner.
drain_queue.py drain_queue.py — the OPERATOR's queue-drainer cockpit.
drain_queue_selftest.py drain_queue_selftest.py — proof harness for the autonomous queue-drainer.
finding_reconcile.py finding_reconcile.py — 2L register hygiene + open-findings digest (Decision 054 follow-on).
firm_analysis_pipeline.py FirmAnalysisPipeline — end-to-end firm analysis across all layers [D4].
generate_model_card_stubs.py Generate model card stubs from model_registry.yaml — Decision 023 Phase 5 / B-07.
generate_model_cards.py Generate automated model cards for all Tier 1 engines [E3].
generate_posture_catalog.py Generate the strategic posture catalog from Decision frontmatter.
heartbeat_health.py WS7 Layer 5 — heartbeat-side health check.
lifecycle_dispatch.py lifecycle_dispatch.py — the model-lifecycle RECONCILER (Decision 054).
merge_clean_prs.py merge_clean_prs.py — the OPERATOR's clean-PR merge cockpit.
model_doc_currency.py model_doc_currency.py — model-documentation currency detector + stamper.
mrm_findings_to_issues.py mrm_findings_to_issues.py — the 2L → git flow mirror (Decision 054, charter §3).
mrm_findings_to_issues_selftest.py mrm_findings_to_issues_selftest.py — proof harness for the 2L→git mirror's
mrm_ratifications_audit.py mrm_ratifications_audit.py — Bridge audit for the 1L↔2L workspace boundary.
mrm_registry_sync.py mrm_registry_sync.py — Phase 0c of the closure-loop pipeline.
observe_portal.py observe_portal.py — the observe-loop probe (G0 / Decision 061).
phase2_lifecycle_schema.py phase2_lifecycle_schema.py — one-off Phase 2 migration
producer_preflight.py producer_preflight.py — make "an autonomous producer opens a PR that fails its
producer_preflight_selftest.py producer_preflight_selftest.py — proof harness for scripts/producer_preflight.py.
quality_engineering_collect.py Parse Playwright JSON results and write quality-engineering agent_state.
rebuild_realized_stats.py Rebuild the DS-068 realized-statistics Derived series from canonical closes.
reconcile_findings.py reconcile_findings.py — reconcile an agent's findings against the principal-owned
reconcile_registry.py reconcile_registry.py — fold per-model registry deltas into the central registries.
reconcile_registry_selftest.py reconcile_registry_selftest.py — proof harness for the registry-delta reconciler.
reconciliation_report.py Generate periodic reconciliation report (D029 Phase 9.2).
security_runtime.py Security runtime checks — runs on the VM via GHA self-hosted runner.
sre_telemetry.py SRE telemetry collector — runs on the VM via GHA self-hosted runner.
sync_ratifications.py sync_ratifications.py — apply the 2L ratification verdict back to the registry.
sync_ratifications_selftest.py sync_ratifications_selftest.py — proof harness for the post-RAT reconciler.
test_auto_rebase_stale_prs.py Selftest for auto_rebase_stale_prs.py — proves the selector picks the right PRs
test_mrm_audit_rule3_under_development.py test_mrm_audit_rule3_under_development.py — the event-based-expiry carve-out.
test_observe_portal_repo_red.py test_observe_portal_repo_red.py — the consumer side of the #3601 repo-red fix.
track_c_crosscheck.py Track C pricing cross-check: FinView vs Gold Copy [C3].
unified_edgar_pipeline.py Unified EDGAR filing data pipeline [D3].
use_case_flows.py Use-case validation flows — runs on the VM via GHA self-hosted runner.
verify_build_claims.py Machine-derived facts for the 1L workflows' Verify gates (#2576).