Read-only auditor for arming-checklist §4 anomalies, the instrument for the
72h-clean finish-line clock. 5 checks against CH + VST (all read-only):
(a) venue order without matching BRIDGE exec_journal row
(b) exec_journal row without venue order
(c) venue order without 'u-' clientOrderId prefix
(d) CH write outside dolphin_uv.* in runner window
(e) tripwire_ok=false occurrences
Authored by cmd-PASS1.2 (PASS8), provenance /mnt/vp-PASS8 bf55777
(agent/oa-violetPASS8). Landed by content (cross-clone fetch hung); files
SHA256-identical to source. Fable verification: 20/20 pass; mutation litmus
RED (inverting the check-(c) prefix guard fails 6 tests incl. PASS8's own
test_c_mutation_* guards — the doctrine's litmus baked into the suite);
read-only confirmed (no writes on any path).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Per Fable #f0fb09ff090feb1f: corrected UTC boundary (gate 08:43:47 CET =
06:43:47 UTC). Re-sliced cached pull by journal_ts. Verified: era A
(incident tail 06:00-06:43:47 UTC) holds 1 of 180 BRIDGE rows (0 pairs,
NET +0.00); era B (clean/certified) holds all 75 paired trades + all 25
bridged-no-venue rows, NET -758.48. Fable's hypothesis (era A carries most
of -758) INVERTED — the full -758.48 is the certified-era number. 0 of 25
no-venue rows are incident-era; they are a clean-era reconciliation gap.
No venue re-fetch. Rollback tag rollback_pre_PASS10_audit_era_20260711190153.
Completes the trade-audit friction half omp could not reach (his sandbox
had no BingX VST DNS). Signed read-only BingX VST pulls (allOrders +
user/income) joined against dolphin_uv.exec_journal BRIDGE rows since
2026-07-10 06:00 UTC. 75 paired SHORT trades, 25 bridged-no-venue-fill gaps.
NET-of-fees P&L = -758.48 USDT (gross -565.35 - fees 198.16 + funding +5.03).
Read-only; VST sandbox only. Rollback tag in /root/violet.git.
- pi_wake_agent.skill.yaml: Full skill definition with all capabilities, parameters, examples, error handling, testing
- pi_wake_agent.skill.json: OpenAI function calling compatible schema
- --dry-run shows what would be done without executing
- Logs intended actions to log file
- Exits 0 without modifying crontab or sending wake messages
NO content authored here — this commits the long-uncommitted LIVE hotfix from
incident 2026-06-12 (bars_held UInt16 poison jammed 18M rows 1.5d): poison rows
retried individually after CH_POISON_ATTEMPTS, then quarantined to dead_letter;
includes the ids loop-variable bugfix. Uncommitted live code was one checkout
away from loss.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Heals reverse vendor drift: the ASEx AccountProjectionV2 adapter was committed on
the vendored copy (violet.git main 824c5cf) instead of upstream, violating the
edit-upstream-then-sync doctrine. Test was orphaned on the PASS9 worktree.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Upstream-canonical commit of the ~366 lines that sat uncommitted in the working
tree (surveyed in DITAV2_SOA_SURVEY_20260702). Venue telemetry published at every
venue boundary (submit/cancel/reconcile) over the zinc venue_region; SOA per
UV DITAv2 adjudication 2026-07-03.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Report path: prod/reports/violet_dark_clickhouse_preclear_20260618T135531Z.md
h5i object capture was skipped because h5i object ingestion was hanging; the report remains at the path above.
PASS 4 completed on worktree agent/oa-violet4 (/mnt/vp-oa4). Recorded in the review queue the two
caveats the OA agent reported — also flagged in-code as prominent banner comments + empty marker
commits on that branch:
- Task 17 exec_intent: explicit reference_price required (ShadowDecision has notional, no price).
- Task 18 alpha_data_feed: parse/feed-port only, no live NT/Binance in unit code.
For Claude's later PASS-4 review (verify the caveats + non-vacuous tests).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
VIOLET_PART_SPEC_OA_TODO_PASS8.md — the V6 bible-layer remainder (ACB done; vol = PASS-4 §14;
sub-second SL guard = PASS-3 §10), six independent units extending contracts_v3 (Posture/
PostureEffect/MarasFingerprint/RegimeView/ActionCadenceDelta):
37. Regime/posture contracts (Posture literal == sizing.Posture).
38. Posture effects engine — 5-state entry-gate/flatten/size-cap, transcribed from the
orchestrator (HIBERNATE_HALT :365, STALKER 2.0 :613, TURTLE regime_dd_halt).
39. MARAS fingerprint consumer — role determined FROM CODE (modulating vs observability-only);
no invented effect.
40. Regime composite read-model (RegimeView; observability/divergence, NOT a sizing path).
41. Cadence shadow-actuation telemetry harness — evaluate-at-fastest / actuate-at-Q deltas (the
shadow evidence required before any Q loosening).
42. Posture/regime parity @gate vs BLUE (bit-identity of gating).
Framing: VIOLET CONSUMES posture/MARAS (BLUE's MHS owns them) — these units APPLY published
effects, they do not run the meta-health state machine. Faithful to the kernel; cite file:line; the
STALKER size-cap stays in sizing.py (this pass owns the entry-gate + observability).
This is the last bible-layer pass; PASS 9 (economics/observability completeness) remains. Added
PASS 8 to the review queue in VIOLET_TODO_CRITICAL.md.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
VIOLET_PART_SPEC_OA_TODO_PASS7.md — the V5 ladder layer as six independent units extending
contracts_v3 (AssetRank/SlotState/CapitalAllocation/SlotPolicy):
31. Selection contracts + SlotPolicy.
32. ARS ranking pipeline (faithful to AlphaAssetSelector.rank_assets + OB Sub-1 ±5/10% + IRP
alignment filter + stablecoin exclusion).
33. Multi-asset slot manager (max_slots, assignment by rank, hysteresis anti-churn, no double-occ).
34. Capital allocation across slots (BLUE notional = base_fraction × conviction × capital; no
invented normalization).
35. Selection→slot→intent multi-asset DARK flow harness.
36. Ranking bit-identity @gate vs the real AlphaAssetSelector.
NOTE (CRITICAL #1): the disappointing live parity (pick-match 1.5%, no-pick 86%, sizing near-
identical when aligned) implicates THIS selection/timing layer. PASS 7 PINS the ranking math
bit-for-bit (Task 36), which narrows the live-parity suspect to timing/join — but the
live-aggregate root-cause stays Claude's job (not this pass). Faithful to AlphaAssetSelector /
BIBLE §5; cite kernel file:line. New-file-only under selection/.
Standing ready for PASS 8–9. Added PASS 7 to the review queue in VIOLET_TODO_CRITICAL.md.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
VIOLET_PART_SPEC_OA_TODO_PASS6.md — the execution internals on top of PASS-5's mock, plus the
quirk-injection layer that flips PASS-5's QuirkProfile seams ON and proves each handler survives.
Six independent units extending contracts_v3 (AccountFrame/PositionFrame/ReconcileCorrection/
OwnershipPolicy):
25. Account/position frame contracts + ownership predicate (foreign-fill filter basis).
26. Fill-pump (ownership filter [quirk #2], dedup, idempotent) → ledger deltas.
27. Reconcile loop (zero-wb guard [#1], reseed-on-update, settle/funding event_seq order [#5]).
28. TTL requote / cancel-replace (+ setLeverage ordering [#7]) over PASS-5 venue + V2 driver.
29. Orphan detection (ORPHAN_VENUE adopt, ORPHAN_LOCAL quarantine — never silent flatten/resurrect).
30. Quirk-injection suite + @gate: flip every QuirkProfile flag ON, assert each is neutralized.
VIOLET's exec reconcile/quirk-handling mirrors PINK's production-tested logic (pink_direct.py is
the authoritative reference — transcribe + cite). New-file-only: quirk injection plugs into PASS-5's
existing seams via NEW fill/frame models, never editing the PASS-5 mock. Real-key boundary smoke
remains MANDATORY before V4-live (injection proves logic, not undocumented venue reality).
Standing ready for PASS 7–9. Added PASS 6 to the review queue in VIOLET_TODO_CRITICAL.md.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
VIOLET_PART_SPEC_OA_TODO_PASS5.md — the DARK mock-BingX execution stack so the full order
lifecycle is buildable/testable with zero keys/risk (V4-live later = swap mock for real client).
Five independent units extending contracts_v3 (Order/OrderAck/Fill/OrderStatus/PositionDelta):
20. Execution contracts + QuirkProfile seam registry (no quirk logic).
21. Mock BingX order FSM (legal lifecycle, partials, reject, reduce-only-increase rejection).
22. Mock BingX venue adapter (port-conformant, injectable fill model, quirk hooks INERT by default).
23. Fill→position/PnL reducer (PnL from fill_price ONLY; ×leverage-aware) → PositionDelta.
24. DARK E2E gate: ExecIntent → router/driver → mock venue → fill reducer → PASS-4 ledger.
CRITICAL per operator: a prominent ⚠️ section makes explicit that BingX "quirks" (zero-wb WS
frames, ownership/foreign-fill collision, bound-price poison, ×leverage notional, settle desync,
reduce-only edge, setLeverage race, dead .pro TLS) are EXPLICITLY OUT OF SCOPE for PASS 5 — the
mock carries injection SEAMS (QuirkProfile, default OFF) to accommodate them LATER, and a real-key
boundary smoke remains MANDATORY before V4-live. Quirks sourced from PINK orphan/reconcile +
DITAv2 audit memory.
Standing ready to write PASS 6–9 (exec internals/reconcile, V5 selection+slots, V6 bible
consumers, economics/observability). Added PASS 5 to the review queue in VIOLET_TODO_CRITICAL.md.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
VIOLET_PART_SPEC_OA_TODO_PASS2.md — four new-file-only tasks that advance the main V0→V6 ladder
toward the DARK-soak gate and V4, each painstakingly specified (why/spec-link, exact new files,
approach, tests, pass criteria), under the same HARD RULES + COMMIT/BRANCH policy as PASS 1:
4. Live BLUE↔VIOLET decision parity report (V3d → soak-readiness gate; reads CH, no writes).
5. Full-DecisionEngine reactor latency gate (V0/V3 budget proof with the real 5-factor brain).
6. L3 tradeability projector standalone (V4 prep; ShadowDecision → exchange-leverage projection,
DARK, not wired to venue/journal).
7. Deterministic replay + golden gate (V3d determinism criterion; same scans ⇒ same decisions).
Independent/parallelizable; recommended order 5→7→4→6. None need the operator's soak/keys
greenlight. PASS-1 tasks reported done — review deferred (noted).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
