Skip to content

Security: Alike001/agentpay-casper

Security

SECURITY.md

Security Policy

Scope

AgentPay Casper is a hackathon prototype deployed for Casper Testnet review.

In scope:

  • Node API and static web console.
  • MCP-compatible JSON-RPC tool gateway.
  • Deterministic policy engine.
  • Odra ReceiptLedger contract and Testnet receipt proof.

Out of scope:

  • Mainnet custody.
  • Production x402 settlement.
  • Production wallet signing infrastructure.
  • Third-party API seller security.

Core Security Model

AgentPay does not make the AI agent trusted. It makes agent actions bounded, inspectable, and revocable.

  • The LLM must not hold private keys.
  • The LLM must not sign transactions.
  • The LLM must not make the final allow/block decision.
  • Policy checks happen before payment execution.
  • On-chain receipts should commit to hashes and payment facts, not raw prompts, secrets, PII, or private API responses.

Reporting a Vulnerability

Open a GitHub issue with the security label if the issue is safe to disclose. For sensitive reports, contact the repository owner privately through GitHub.

Please include:

  • Affected route, package, or contract.
  • Reproduction steps.
  • Expected vs actual behavior.
  • Severity estimate.

Known Limitations

  • Testnet prototype only.
  • Contracts are unaudited.
  • Not suitable for custody or mainnet funds.
  • The current x402 flow is an x402-style prototype for hackathon review.

There aren't any published security advisories