purushowtham/Algo-proj ? reverse-engineered prompt

Build me a TypeScript backend called AgentVault that sits in front of AI agent requests and keeps spending under control when the agent needs to use paid x402 services.

I want it to check a target URL first, block unsafe or private destinations, figure out whether the request needs payment, and then decide if the agent is allowed to continue based on simple rules like allowed and blocked domains, daily budget, per request budget, and a basic trust score. If the request is approved, it should reserve the spend before execution, complete the payment flow, and save a receipt so I can audit what happened later.

Please include the API routes for health, agent creation and lookup, spending status, inspect, authorize, execute, reputation check, router resolve, router execute, and receipt lookup. Use SQLite locally, Prisma for storage, and make it work with the local x402 and Algorand Testnet setup from the README. Keep it runnable with a simple dev workflow, and add tests for the main approval and payment flows. If you need current docs for any x402 or Algorand details, look them up online.