Contract Skill — A ready-to-use MOVA HITL workflow. Requires the openclaw-mova plugin.

MOVA Credit Scoring

Submit a loan application to MOVA for automated credit risk scoring — with explainable risk band, bureau check, and a mandatory human credit officer decision gate backed by a full audit trail.

What it does

1. 1. Risk scoring — AI evaluates income, debt-to-income ratio, bureau score, and repayment history against the scoring model
2. Risk band — applicant assigned a risk band (excellent / good / fair / poor / very_poor) with score 0–1000
3. Credit limit recommendation — AI suggests approved amount based on risk band and requested amount
4. Human gate — credit officer reviews the scoring breakdown and chooses: approve / approve at reduced limit / reject / request more info
5. Audit receipt — model version, all input features, the human identity, and the decision timestamp are logged for regulatory accountability

Mandatory escalation rules enforced by policy:

• - Risk band poor or verypoor → mandatory human review, cannot auto-approve
• Requested amount above threshold → always routes to human gate
• Bureau score missing or frozen → requestinfo required, no auto-decision

Requirements

Plugin: MOVA OpenClaw plugin must be installed in your OpenClaw workspace.

Data flows:

• - Applicant ID + financial data → api.mova-lab.eu (MOVA platform, EU-hosted)
• Financial features → credit scoring model (server-side, read-only)
• Applicant ID → credit bureau (read-only, no data stored)
• Audit journal → MOVA R2 storage, signed
• No data sent to third parties beyond the above

Demo

Step 1 — Application submitted: APP-2026-0041, €25K, bureau score 612

Step 2 — AI scoring: risk band fair, DTI 0.017, approve_reduced recommended at €5K

Step 3 — Decision recorded: approve_reduced + audit receipt

Quick start

Say "score credit application APP-2026-0041 for applicant CUST-1501" and provide:

CODEBLOCK0

The agent submits the application, shows the AI risk band and score breakdown, then asks for the credit officer's decision.

Why contract execution matters

• - Scoring rules are policy, not prompts — poor risk bands and high amounts trigger mandatory gates that cannot be bypassed
• Full explainability — every contributing factor (income, debt ratio, bureau score) is surfaced with its weight
• Immutable audit trail — when a regulator or applicant challenges a rejection, the complete scoring chain with timestamps is in the system
• EU AI Act / CRD VI / EBA GL ready — credit decisions are high-risk AI outputs requiring human oversight, explainability, and a documented decision chain

What the user receives

Output	Description
Credit score	0–1000 numerical score
Risk band

excellent / good / fair / poor / very_poor | | Debt-to-income ratio | Calculated from input data | | Bureau result | Bureau score + credit history summary | | Anomaly flags | highdti, lowbureauscore, shorthistory, missing_bureau | | Findings | Structured list with severity codes | | Recommended credit limit | AI-suggested approved amount | | Recommended action | AI-suggested decision | | Decision options | approve / approvereduced / reject / requestinfo | | Audit receipt ID | Permanent signed record of the credit decision | | Compact journal | Full event log: scoring → bureau check → human decision |

When to trigger

Activate when the user:

• - Mentions an application ID or loan request (e.g. "APP-2026-0041")
• Says "score credit", "credit assessment", "loan decision", "assess borrower", "credit risk review"
• Provides applicant financial data for a lending decision

Before starting, confirm: "Submit application [APP-ID] for MOVA credit scoring?"

If applicant ID or requested amount is missing — ask once.

Step 1 — Submit loan application

Call tool mova_hitl_start_credit with:

• - application_id: application reference (e.g. APP-2026-0041)
• INLINECODE4: customer/borrower ID
• INLINECODE5: loan amount requested
• INLINECODE6: applicant annual income
• INLINECODE7: existing monthly debt obligations
• INLINECODE8: credit bureau score (optional)
• INLINECODE9: employed / self_employed / unemployed / retired
• INLINECODE10: optional (mortgage, personal, auto, business)

Step 2 — Show scoring result and decision options

If status = "waiting_human" — show risk scoring summary and ask to choose:

CODEBLOCK1

Option	Description
INLINECODE12	Approve at requested amount
INLINECODE13

Approve at reduced credit limit |
| reject | Reject application |
| request_info | Request additional financial information |

Call tool mova_hitl_decide with:

• - contract_id: from the response above — this is ctr-crd-xxxxxxxx, NOT the application ID
• INLINECODE19: chosen decision
• INLINECODE20: credit officer reasoning

Step 3 — Show audit receipt

Call tool mova_hitl_audit with contract_id.
Call tool mova_hitl_audit_compact with contract_id for the full signed scoring chain.

Connect your real credit systems

By default MOVA uses a sandbox mock. To route scoring against your live infrastructure, call mova_list_connectors with keyword: "credit".

Relevant connectors:

Connector ID	What it covers
INLINECODE27	Internal credit scoring model
INLINECODE28

External credit bureau score and history |

Call mova_register_connector with connector_id, endpoint, optional auth_header and auth_value.

Rules

• - NEVER make HTTP requests manually
• NEVER invent or simulate credit scores — if a tool call fails, show the exact error
• Use MOVA plugin tools directly — do NOT use exec or shell
• CONTRACTID is ctr-crd-xxxxxxxx from the movahitlstartcredit response — NOT the application ID