Select Workflow
Choose one area: refunds, payment retries, treasury, stablecoin liquidity, collateral, wallet policy, limits, reporting, or evidence from existing financial-risk tools.
Open guideEvaluation Flow
The first financial runtime review should test one workflow family with synthetic or sanitized metadata. Existing financial controls remain authoritative.
Choose one area: refunds, payment retries, treasury, stablecoin liquidity, collateral, wallet policy, limits, reporting, or evidence from existing financial-risk tools.
Open guideProvide 5 to 25 action records with risk, reversibility, containment, rollback latency, evidence, anomaly pressure, and scope.
Open sampleUse the repository workflow from GitHub Actions with read-only repository permissions.
Run public workflowCompare financial runtime posture, execution route, controls, Ref-gate failures, ledger validity, autonomy-budget impact, and any external risk signals used as evidence.
Open reportWhat The Sample Shows
The sample intentionally mixes actions that can proceed, actions that should be constrained, and actions that should be blocked because recovery is weak or hard evidence fails. The point is not to detect crime or certify compliance. The point is to test whether automated financial actions deserve different runtime postures before execution.
The finance pack reuses the general customer-evaluation runner with a financial runtime sample file. The internal profile name is SMERC-F.
python -m reference_engine.customer_evaluation \
examples/smerc_f_customer_eval_actions.json \
--json-output reports/smerc_f_customer_evaluation/customer_evaluation_report.json \
--markdown-output reports/smerc_f_customer_evaluation/Customer_Evaluation_Report.md \
--pretty
Move forward only if a financial-services reviewer can provide 10 to 25 metadata-only actions from one workflow, label whether each posture is useful or too strict, and define success metrics for a shadow-mode review. Existing financial controls should remain the source of truth during the first pilot.