HK Snapshot Monthly Audit: 2026-07

[github-actions]

HK Snapshot Monthly Audit: 2026-07

This issue is the monthly AI audit input for HK snapshot-backed strategy scaffolds.
It is intended for CodexAuditBridge and does not publish artifacts, deploy Cloud Run, or place broker orders.

Routing

• Source repo: QuantStrategyLab/HkEquitySnapshotPipelines
• Source ref: main
• Codex audit repo: QuantStrategyLab/CodexAuditBridge
• Codex task: monthly_snapshot_audit
• Bundle version: hk_snapshot_monthly_audit_bundle.v1

Active snapshot candidate

Rank	Profile	Stage	Required drawdown gate	Next evidence focus
1	hk_low_vol_dividend_quality_snapshot	production_data_audit_and_walk_forward_first	max drawdown <= 30%	Audit production snapshot source, then run survivorship-safe walk-forward backtests.

This profile remains an architecture_scaffold candidate until the evidence gates below pass.
No live-enable approval should be inferred from sample artifacts or from this monthly audit issue alone.

Non-selected snapshot profiles

• hk_shareholder_yield_quality — research-only / deprioritized until stronger validated evidence exists.
• hk_free_cash_flow_quality — research-only / deprioritized until stronger validated evidence exists.
• hk_residual_momentum_quality — research-only / deprioritized until stronger validated evidence exists.
• hk_factor_mix_qvlm_risk_parity — research-only / deprioritized until stronger validated evidence exists.
• hk_quality_growth_low_volatility — research-only / deprioritized until stronger validated evidence exists.
• hk_southbound_flow_momentum — research-only / deprioritized until stronger validated evidence exists.
• hk_central_soe_value_quality_select — research-only / deprioritized until stronger validated evidence exists.
• hk_composite_factor_quality_value_momentum — research-only / deprioritized until stronger validated evidence exists.
• hk_liquid_momentum_quality — research-only / deprioritized until stronger validated evidence exists.
• hk_blue_chip_leader_rotation — research-only / deprioritized until stronger validated evidence exists.
• hk_ah_premium_relative_value — research-only / deprioritized until stronger validated evidence exists.
• hk_index_rebalance_event — research-only / deprioritized until stronger validated evidence exists.
• future_research_long_tail — research-only / deprioritized until stronger validated evidence exists.

Required gates for AI review

1. sample_artifact_smoke — Build sample artifacts only to verify contract wiring; sample artifacts are not live data.
2. production_source_audit — Replace sample inputs with point-in-time production data and document source coverage, quality reports, corporate actions, suspensions, and missingness controls.
3. walk_forward_backtest — Run survivorship-safe walk-forward backtests with at least three independent OOS folds, net HK costs, and max drawdown at or below the profile gate.
4. artifact_pack_validation — Validate the published snapshot CSV, manifest, ranking, release summary, and provenance.
5. platform_dry_run_evidence — Run IBKR and LongBridge dry-run order previews with lot-size, liquidity, fee, quote, bilingual notification, and delivery-log evidence.
6. operator_approval_and_rollout_plan — Collect approval reference, staged rollout plan, tripwires, kill switch, and rollback plan.

The backtest gate requires point-in-time inputs, no look-ahead bias, no survivorship bias, at least 3 independent OOS folds, and max drawdown <= 30% after HK costs.

The dry-run gate requires LongBridge and IBKR order-preview evidence with lot-size, liquidity, fee, quote, capacity, and bilingual notification evidence. Operator approval is still required before any live-enable change.

Operator checklist for this month

• Confirm whether production factor snapshots replaced all sample inputs.
• Confirm artifact pack validation outputs and immutable evidence URIs are attached.
• Confirm walk-forward backtests remain net of HK costs/slippage and within the 30% drawdown cap.
• Confirm dry-run order previews exist for both LongBridge and IBKR before promotion.
• Confirm English and Chinese notification previews/delivery logs use the unified notification format.

Reproducible local commands

python scripts/print_first_snapshot_promotion_plan.py --json
python scripts/print_snapshot_promotion_matrix.py --json
python scripts/print_snapshot_readiness.py --profile hk_low_vol_dividend_quality_snapshot --platform longbridge --json
hkeq-validate-live-enable-evidence --evidence-file <live-enable-evidence.json> --json

Snapshot matrix summary

• Matrix status: architecture_scaffold_not_live_enabled
• Source project: HkEquitySnapshotPipelines
• Active candidates: hk_low_vol_dividend_quality_snapshot
• Live enablement without evidence: False

Requested Codex output

Please provide a concise bilingual review covering release consistency, evidence gaps, downstream impact, and operator action items. Do not recommend production trading changes unless all evidence gates are present.

[quantcrossrepoautomation]

Codex Audit

CodexAuditBridge stopped before making repository changes because the service backend failed outside the source PR/test surface.

• Failure category: auth_or_config_failure
• Detail: Codex audit service request failed: 405 {"status":"error","error":"method not allowed"}

No files were pushed and no PR feedback retry is needed for this run.

[quantcrossrepoautomation]

Codex Audit

CodexAuditBridge stopped before making repository changes because the service backend failed outside the source PR/test surface.

• Failure category: auth_or_config_failure
• Detail: Codex audit service request failed: 405 {"status":"error","error":"method not allowed"}

No files were pushed and no PR feedback retry is needed for this run.

[quantcrossrepoautomation]

Codex Audit

CodexAuditBridge stopped before making repository changes because the service backend failed outside the source PR/test surface.

• Failure category: transient_service_failure
• Detail: The read operation timed out

No files were pushed and no PR feedback retry is needed for this run.

[quantcrossrepoautomation]

Codex Audit

Verdict

Blocked. I fixed a low-risk monthly-report defect, but issue #38 still lacks the evidence required for HK snapshot promotion, and the generated staging input path remains unsafe as production-source evidence.

Findings

• High: src/hk_equity_snapshot_pipelines/low_vol_dividend_longbridge_snapshot.py and src/hk_equity_snapshot_pipelines/low_vol_dividend_public_snapshot.py derive non-price fields such as market_cap_hkd, payout_ratio, earnings_positive, pe_ratio, and free_cash_flow_yield from provider valuation lookups that do not accept an as_of timestamp. For backdated runs this creates look-ahead risk, so public_yfinance_staging / longbridge_openapi_staging should not be treated as point-in-time production-source audit evidence without human follow-up.
• Medium: src/hk_equity_snapshot_pipelines/monthly_snapshot_audit_issue.py only emitted a longbridge readiness command even though the monthly gate explicitly requires evidence for both longbridge and ibkr. That could lead operators to run an incomplete monthly checklist. Fixed.

Changes Made

• Updated src/hk_equity_snapshot_pipelines/monthly_snapshot_audit_issue.py so the generated monthly issue now prints reproducible print_snapshot_readiness.py commands for both longbridge and ibkr.
• Added tests/test_monthly_snapshot_audit_issue.py to lock the dual-platform readiness commands into the monthly issue payload.

Verification

• Added focused regression coverage for the monthly issue payload.
• Repository-local command execution could not be run in this sandbox because every shell attempt failed with bwrap: loopback: Failed RTM_NEWADDR: Operation not permitted.

Residual Risk

• Issue HK Snapshot Monthly Audit: 2026-07 #38 still provides no point-in-time production-source audit, no 3-fold OOS backtest pack, no dual-broker dry-run evidence, no bilingual notification delivery-log evidence, and no operator approval pack; promotion must remain blocked.
• The staging snapshot generators still need a human decision: either downgrade them to research-only evidence or add audited point-in-time fundamentals plus availability-date lineage before operators use them for monthly promotion evidence.

Created fix PR: #39

This PR is high-risk for unattended merge and requires human review.

• Risk level: high
• Changed lines: 407
• Reasons:
  • invalid auto-merge policy schema requires human review
• High-risk files:
  • src/hk_equity_snapshot_pipelines/monthly_snapshot_audit_issue.py
  • tests/test_monthly_snapshot_audit_issue.py

Added human-review-required to the PR so operators can review it before merge.