Pre-Flight Audit — Deploy + Verify Phase 0
Closes: V-001, V-002, V-003, V-004, V-005 in Deploy + Verify Tracker §2.
Discipline applied: SUBAGENT_ORCHESTRATION_DISCIPLINE_2026-05-16.md §7 (existing-program search before dispatch) + §2 (orchestrator verification per claim).
V-001: Branch inventory (pushed vs local)
Before this audit: 2 of 18 parity branches on origin. After RESCUE PUSH: 18 of 18 parity branches on origin. Data-loss risk eliminated.
Parity-task branches on fin-platform/fin-agentic origin (post-rescue)
| Branch | Subagent | Status |
|---|---|---|
| claude/parity-task-A-01-A-04-stock-parity | A (ac5ce5c1) | PUSHED ✓ |
| claude/parity-task-C-01-C-05-renderers | C (a91b09b3) | PUSHED ✓ |
| claude/parity-task-D-01-D-06-proof-qc-photos | D (a11914bb) | PUSHED ✓ |
| claude/parity-task-E-01-E-08-projection-audit | E (a3889ff0) | PUSHED ✓ |
| claude/parity-task-F-01-F-06-frontmatter | F (af830706) | PUSHED ✓ |
| claude/parity-task-G-01-G-09-agentic-hardening | G (a8975b2c) | PUSHED ✓ |
| claude/parity-task-H-01-H-06-memory-opt | H (a5addf03) | PUSHED ✓ |
| claude/parity-task-I-01-I-10-briefings | I (ab495aa1) | PUSHED ✓ |
| claude/parity-task-J-mcp-projections | J (a6a37241) | PUSHED ✓ |
| claude/parity-task-K-01-K-08-team-profiles | K (aad933ae) | PUSHED ✓ (rescued after worktree cleanup) |
| claude/parity-task-L-01-L-03-teams-importers | L day-1 (a334ed8c) | PUSHED ✓ |
| claude/parity-task-L-04-L-08-historical-remainder | L day-3 (acbf1ad9) | PUSHED ✓ |
| claude/parity-task-M-01-M-07-fusion-engine | M.1 (a9a7938d) | PUSHED ✓ |
| claude/parity-task-N-01-N-11-drill-throughs | N (a838c6ab) | PUSHED ✓ |
| claude/parity-task-N-01-drill-design | N vault sibling | PUSHED ✓ |
| claude/parity-task-P-01-P-09-data-discipline | P (a5513eb0) | PUSHED ✓ |
| claude/parity-task-P-catalog-docs | P vault sibling | PUSHED ✓ |
| claude/parity-task-day5-final-wave | Day 5 (ae0707d6) | PUSHED ✓ |
Vault branches on fin-platform/fin-central-intranet
| Branch | Subagent | Contains |
|---|---|---|
| origin/main (current) | (orchestrator merges) | 101 commodities + 142 vendors + 244 varieties + 9 audit/design docs + reconciled tracker |
| claude/parity-task-A-04-stock-parity | A | 14 starter commodity profiles + audit |
| claude/parity-task-F-04-vault-mirror | F | 357 frontmatter additions + 4 new canonical docs + spec amendment |
| claude/parity-task-C-06-vendor-bulk | C | 142 vendor profile docs |
| claude/parity-task-day5-w19-apex | Day 5 | W19 proof package (synthetic) + .pdf |
Integration branch status
| Branch | Status |
|---|---|
claude/parity-program-integration on fin-agentic | ❌ NEVER CREATED. Auto-merge workflow has no target. Required for V-101. |
V-002: Subagent worktrees inventory
/Users/robert/Projects/FIN/_03_Dev/fin-agentic/.claude/worktrees/agent-*
13+ subagent worktrees from this session. Most are still on disk. K + P-catalog-docs worktrees were cleaned up before push (branches rescued via main repo ref).
V-003: Migration collision map
| Slot | Branch A | Branch B | Resolution |
|---|---|---|---|
| 049 | claude/parity-task-day5-final-wave (doc_backlink_projection) + claude/heuristic-kepler-4f9104 (applied to fin-central-prod) | claude/iar-rm-033-lot-lifecycle-reconciler-20260515 (lot_lifecycle_reconciliation_projection) | doc_backlink wins (already applied to fin-central-prod). IAR-RM-033 must renumber to 064+ at its eventual merge. Documented in E-FU-1. |
| 050 | doc-CQRS (entity_doc_projection — applied to fin-central-prod) | — | Clean |
| 051 | Part H (actor.memory JSONB indexes — applied) | — | Clean |
| 052 | Part H (operational_memory archival — applied) | — | Clean |
| 053 | Part N (drill_pattern_projection — applied) | — | Clean |
| 054 | Part D (proof_package_projection — applied) | — | Clean |
| 055 | Part J (customer_360 — NOT applied) | Part G subagent renumbered duplicate of doc_backlink_projection | Drop G’s duplicate; keep J |
| 056 | Part J (vendor_360 — NOT applied) | Part G subagent renumbered duplicate of entity_doc_projection | Drop G’s duplicate; keep J |
| 057 | Part J (commodity_360 — NOT applied) | — | Clean |
| 058 | Part J (location_intel — APPLIED to fin-central-prod) | — | Clean |
| 059 | Part J (supply_chain_position — NOT applied) | — | Clean |
| 060 | Part J (daily_company_pulse — NOT applied) | — | Clean |
| 061 | Part J (sales_team_dashboard — NOT applied) | — | Clean |
| 062 | Part J (buyer_team_dashboard — NOT applied) | — | Clean |
| 063 | Part P (data_quality_dashboard — NOT applied) | — | Clean |
Action: When merging integration branch, drop Part G’s duplicate 055/056 migration files (G’s renderer code stays; just the migration files conflict with J).
V-004: Baseline event/projection counts
fin-central-prod (Neon project cool-scene-51259109)
| Metric | Value |
|---|---|
events count | 0 (empty by design — Central control plane, not tenant data) |
knowledge count | 1 (FSMA 204 seed) |
| schema version | v58 (049, 050, 051, 052, 053, 054, 058 applied) |
| Projection tables live | 11 (041-044 + 049/050/051/052/053/054/058) |
nathel-prod (Neon project wispy-mud-49174151)
⚠ CRITICAL SCOPE CORRECTION 2026-05-16T23:50Z (per Rob): Up until ~7 days ago, all operational data in nathel-prod was purely testing / synthetic seed / scaffolding. The 12.1M total is dominated by auth.login (5.6M), auth.actor_mapped (5.5M), security.rate_limit (534K), daemon.heartbeat (256K) — i.e., test infrastructure noise.
REAL data sources (the only ones to use for verification):
-
Email intake via
intel@fintail.net+arrivals@fintail.net— events since ~2026-05-09 (when Rob’s data-spine fixes landed). Roughly:- 203
signal.email_extract(real email body extractions) - 199
extractor.correspondence_extracted - 203
extractor.dispatch_planned - 80
signal.pp_arrival_logged(real RP581 arrival sheets from arrivals@) - 73
signal.po_received/ 47signal.po_created - 17
extractor.routing_decided - 5
ingest.pdf_extracted(extractor bottleneck — Rob’s IAR-RM-026/027 work resolving) - 4
bridge.report_imported(real PP report imports) - 404
channel.message_received/channel.message_routed(underlying transport) - 49
lot.pre_arrival_synthesized - 174
ingest.intel_pdf_classified - Total real-data signal events: ~400-600 (NOT millions)
- Date range: 2026-05-09 → present (LIVE, growing)
- 203
-
Teams + QC channel data — historical real conversations:
- 146 messages in
/Users/robert/Projects/FIN/_03_Dev/teams-qc-export-2026-05-06.json - Glenn Jones 50 day-shift narratives + Anthony Napolitano 50 night-shift + Pedro 15 coordinator + Joshua 10 strategy memos + Keith Martin 6 salesman + Mike Donovan 6 CA FOB
- Plus 110 narrative↔photo pairings in
teams-qc-photo-narrative-pairing-2026-05-06.json - Date range: 2025-01 → 2026-05 (17-month historical corpus)
- NOT yet in events table — Part L importer
--applypath not yet implemented (deferred per Deploy+Verify pivot since Rob’s OPERATIONAL_DATA_SPINE may cover it)
- 146 messages in
ALL OTHER nathel-prod events (the other ~12,096,000) are synthetic / test / scaffolding. DO NOT cite as real Nathel operational data.
Real-data commodity coverage (last 7 days, sample)
Parser still naïve — returns brand prefixes as commodities. Actual coverage:
- 21 ORG (generic organic prefix)
- 18 EB (Earthbound Farms brand prefix)
- 7 ATT (Attitude brand prefix)
- Long tail: ORG BLACKBERRY, ORG BASIL, ORG FRENCH BEAN, ORG PKG MINI CUCUMBER, BELL, SHALLOT, etc. (1-2 events each)
Implication: agentic eval test queries should target commodities/vendors ACTUALLY in the real-data corpus (not arbitrary “Cucumber Hot House” or “Mission Produce” that may not have signals yet). The §10 watermelon test will FAIL until watermelon shows up in real intake — must use a substitution query from the actual real-data subset.
fin-central-prod (Neon project cool-scene-51259109)
| Metric | Value |
|---|---|
events count | 0 (empty by design — Central control plane, not tenant data) |
knowledge count | 1 (FSMA 204 seed) |
| schema version | v58 (049, 050, 051, 052, 053, 054, 058 applied) |
| Projection tables live | 11 (041-044 + 049/050/051/052/053/054/058) |
Verify target scope (BINDING for V-401b onward)
When querying nathel-prod for verification:
WHERE timestamp >= '2026-05-09T00:00:00Z'
AND (type LIKE 'signal.%'
OR type LIKE 'extractor.%'
OR type LIKE 'ingest.%'
OR type LIKE 'bridge.%'
OR type IN ('channel.message_received','channel.message_routed',
'lot.pre_arrival_synthesized'))Plus the Teams export file (separate corpus) when querying historical context.
V-005: Program overlap search (Rule §7 application)
Found ~60 in-flight or recent programs in fin-agentic/docs/processes/ that touch parity-adjacent scope. The ones most overlapping with Deploy + Verify work:
| Program | Date | Overlap with parity Deploy+Verify |
|---|---|---|
| INTERNAL_AGENT_READINESS_PROGRAM_2026-05-12 | 5/12 | Central readiness tracker for IAR-RM-NNN lanes. Should add parity-specific V-lanes here too. |
| OPERATIONAL_DATA_SPINE_AND_ENTITY_INTELLIGENCE_READINESS_PROGRAM_2026-05-14 | 5/14 | ~70% overlap with Deploy + Verify Phase 4 (Seed). 6 lanes already FIXED_VERIFIED (IAR-RM-022..027). Defer intake/sales/RP581/pricing/enrichment to this program. |
| MARKET_INTEL_SIGNAL_OPERATIONALIZATION_REMEDIATION_PROGRAM_2026-05-12 | 5/12 | Overlaps with Part O external sources + Part M.1 pricing fusion. Coordinate before V-501..V-506. |
| EMAIL_INTEL_FUSION_AND_LIVING_SIGNAL_EXECUTION_PROGRAM_2026-05-04 | 5/4 | Parent of intel@/arrivals@ pipeline. Coordinate Part L historical importer scope. |
| PAE_AGENT_QUALITY_LIFT_EXECUTION_PROGRAM_2026-04-28 | 4/28 | Overlaps with Part G agentic hardening + Phase 7 LIVE gradeboard eval. MAJOR overlap — must read before V-701..V-704. |
| DATA_PLANE_OPTIMIZATION_EXECUTION_PROGRAM_2026-04-30 | 4/30 | Overlaps with Part E projection audit + Phase 5 backend verify. |
| KNOWLEDGE_BASE_OPTIMIZATION_EXECUTION_PROGRAM_2026-04-30 | 4/30 | Overlaps with Part B/C entity profile work + Phase 5 backend verify. |
| FINN_ITE_QC_LIVE_TEST_EXECUTION_PROGRAM_2026-05-02 | 5/2 | Overlaps with Phase 6 Playwright (QC inspector role flow). Likely already has JWT mint pattern. |
| CICD_PIPELINE_EFFICIENCY_REMEDIATION_PROGRAM_2026-04-23 | 4/23 | Why Rob said “direct deploy because CI/CD is stale and slow”. |
| NATHEL_FIN_COLLABORATION_AND_REPO_HYGIENE_EXECUTION_PROGRAM_2026-05-12 | 5/12 | Repo hygiene patterns relevant for Phase 1. |
IAR-RM-* branches in flight on master/staging
| Branch | Topic |
|---|---|
claude/iar-rm-018-event-alignment-closure-20260516 | Event alignment closure |
claude/iar-rm-019-capability-decomposition-20260516 | Capability decomposition |
claude/iar-rm-028-pp-date-mirror-20260515 | PP date mirror |
claude/iar-rm-029-qc-mobile-ui-20260515 | QC mobile UI |
claude/iar-rm-029-staging-proof-20260516 | QC mobile UI staging proof |
claude/iar-rm-030-actual-received-at-producer-20260515 | actual_received_at producer |
claude/iar-rm-031-035-arrivals-pipeline-20260515 | Arrivals pipeline (RM-031..035) |
claude/iar-rm-031-invadj-ingestion-20260515 | Inventory adjustment ingestion |
claude/iar-rm-033-lot-lifecycle-reconciler-20260515 | CONFIRMED: source of 049 migration collision |
claude/iar-rm-034-no-load-no-lot-handling-20260515 | No-load-no-lot handling |
claude/iar-rm-035-sales-schema-extension-20260515 | Sales schema extension |
claude/iar-rm-036-inventory-snapshot-ingestion-20260515 | Inventory snapshot ingestion |
claude/iar-rm-038-source-artifact-maintenance-cron-fix-20260516 | Source artifact maintenance cron fix |
claude/master-ci-cleanup-confirm-load-arrival-event-alignment-20260516 | CI cleanup + load arrival alignment |
Action items for Phase 1 (V-101..V-111):
- Coordinate with Rob’s existing OPERATIONAL_DATA_SPINE program — don’t duplicate; defer overlapping lanes
- Read PAE_AGENT_QUALITY_LIFT before V-701..V-704 (Phase 7 LIVE eval) — major overlap; likely defines gradeboard methodology
- Read INTERNAL_AGENT_READINESS_PROGRAM — likely central tracker we should cross-link to
- Schedule IAR-RM-033 renumber when it merges (it will collide with already-applied 049 doc_backlink)
Acceptance — V-001..V-005 closed
| Task | State | Proof |
|---|---|---|
| V-001 | COMPLETED | This doc §V-001 — pre/post rescue counts documented |
| V-002 | COMPLETED | This doc §V-002 — worktree inventory |
| V-003 | COMPLETED | This doc §V-003 — migration collision map |
| V-004 | COMPLETED | This doc §V-004 — baseline counts (verified via Neon MCP queries) |
| V-005 | COMPLETED | This doc itself (vault commit) — docs/audits/PRE_FLIGHT_2026-05-16.md |
Recommended next moves (Phase 1)
Given the program overlap discoveries above, Phase 1 (V-101..V-111) needs minor re-scoping before execution:
- V-101 Create integration branch — proceed as planned
- V-102 Push subagent branches — ALREADY DONE in this audit (18/18 on origin)
- V-103 EntityType duplicate fix — proceed as planned
- V-104..V-109 Merge branches in dep order — proceed, but ALSO coordinate with
claude/iar-rm-033-lot-lifecycle-reconcilerto renumber before its merge - V-110 Local pre-flight — proceed as planned
- V-111 Push integration branch — proceed as planned
Before any Phase 4+ work:
- Read INTERNAL_AGENT_READINESS_PROGRAM_2026-05-12 to register parity V-lanes in the central tracker
- Read PAE_AGENT_QUALITY_LIFT_EXECUTION_PROGRAM_2026-04-28 to align with existing gradeboard methodology (skip duplication for V-701..V-704)
- Confirm with Rob: does he want parity V-lanes in his central tracker, or kept separate in vault?