RWA Credit SentinelFinals v3

Casper Agentic Buildathon 2026 · Final Round

From verified RWA risk to bounded capital action.

An underwriting agent verifies evidence, anchors a risk credential on Casper, and prepares a policy-constrained lending intent with a complete audit trail.

Casper networkTestnet
Policy checks9
BenchmarkLoading

Verified finals evidence

Two real state transitions on Casper

The finals contract stores both the underwriting credential and its bounded capital intent. Every link below is independently verifiable on Testnet.

Qualification contract remains available
Finals deploy3a305efe3c7233…9a458e591cRisk credentialda9174726c74f1…7d632cebd6Execution intent68175104219126…5ebbe2542a
Contract6a248275de2c45…504ed9cd4f
Package2c34005155776d…0a7e98358f
Credential staterecord_credent…aa7fc2ad52
Intent staterecord_executi…d3ea7816db

One click reads both dictionaries from Casper RPC and compares 13 fields with the published finals evidence.

01 / Underwrite

Build the verifiable risk credential

Evidence remains off-chain. Content hashes, the explainable report, and the credit decision become a Casper-verifiable credential.

Evidence normalization Explainable risk scoring Credential anchoring
Financing request
Content-hashed synthetic evidence pack4 documents / invoice register, purchase orders, deliveries, and payment history
Inspect JSON
Evidence intakeJSON, CSV, TXT or PDF / 1 MB each / 6 files maximum

02 / Credential

Consume evidence, not trust

The execution desk only accepts a credential created by this registry workflow. Report and evidence hashes connect the off-chain analysis to Casper state.

Run underwriting to create a credentialComplete the preceding workflow step to continue.

03 / Execute

Set the agent's financial boundaries

The risk agent proposes. Deterministic vault policy disposes. Private keys stay outside the model, and exceptions route to reviewer multisig.

A verified credential is requiredComplete the preceding workflow step to continue.

Evaluation

Safety claims backed by repeatable tests

The benchmark is deterministic by design: it validates policy behavior, while chain links validate Casper integration.

Cases
Decision agreement
Invalid credentials blocked
Policy failure modes