@kashdao/protocol-sdk is non-custodial software by construction. It is also zero-Kash-dependency software: nothing on the trade path reaches a Kash-operated server.

• Kash never holds or controls user funds. USDC, outcome tokens, and any other on-chain assets sit at addresses the customer controls. Kash holds zero balances on the customer's behalf at any time.
• Kash never has access to user signing keys. This SDK accepts a consumer-supplied signer adapter. Keys live wherever the consumer decides — a local viem account, a remote HSM (AWS-KMS, GCP-KMS), Fireblocks, web3signer over RPC, Privy MPC, a hardware wallet — and never transit any Kash-controlled boundary.
• Kash is not on the trade path at all. This SDK runs entirely on the consumer's infrastructure (consumer's RPC, consumer's signer, consumer's bundler). The Kash backend is never contacted; Kash cannot inspect, block, censor, delay, or replay trades made through this SDK.
• Kash never moves user funds. Settlement is on-chain via open-source protocol contracts. Positions are tokenized ERC-1155 outcome tokens held directly by the user's account; there is no Kash-controlled pool, ledger, or relay in the path.
• Kash is not a money-services business, custodian, exchange, or broker-dealer. Kash publishes software and protocol contracts; customers run the software and interact with the protocol from accounts they control.

Equivalent statements hold for @kashdao/sdk, @kashdao/cli, and kashdao-protocol-sdk (Python).

Please do not file a public GitHub issue for security vulnerabilities.

Email: security@kash.bot

Include:

• A clear description of the issue
• Reproduction steps or proof-of-concept
• The version of @kashdao/protocol-sdk (and any relevant runtime info: Node version, browser, viem version, bundler vendor, signer type)
• Whether the issue affects the SDK, the consumer-side smart-account flow, or both
• Your contact info if you'd like attribution in the disclosure

We aim to:

1. Acknowledge receipt within 2 business days.
2. Confirm or reject the report within 7 business days.
3. Ship a fix within 30 days for accepted reports, faster for issues actively being exploited on mainnet.

Once the fix has shipped and a reasonable upgrade window has elapsed, we publish a coordinated disclosure crediting the reporter (with consent).

In scope:

• Vulnerabilities in @kashdao/protocol-sdk itself (this package).
• Issues that allow bypassing the SDK's documented guarantees:
  • UserOp hash divergence — calldata or hash that disagrees with what EntryPoint v0.7 / viem.getUserOperationHash produce, causing signed UserOps to be rejected or accepted incorrectly
  • ABI drift — vendored ABIs that disagree with @kashdao/contract-abis in a way the CI drift test fails to catch
  • Address drift — per-chain address registry that disagrees with @kashdao/protocol-config
  • Signer adapter leakage — any path where the SDK persists, logs, or transmits signed bytes, raw hashes, or signer-side credentials beyond what's required to forward a UserOp to the consumer-configured bundler
  • REST client SSRF / open-redirect via apiUrl validation gaps
  • HTTPS bypass — http:// apiUrl accepted in production paths (currently rejected unless localhost)
  • HTTP smuggling / header injection via config inputs
  • Slippage-protection bypass — calldata that ignores maxSlippageBps or routes around the on-chain quote check
  • Decoder injection — revert-data decoding that allows an attacker to surface forged error names through simulate()

Out of scope (please report to security@kash.bot separately if relevant):

• Vulnerabilities in the Kash protocol contracts themselves (Market, Vault, Oracle, MarketFactory, OutcomeTokens1155, ParamRegistry) — these are in-scope for the protocol bug bounty, not the SDK
• Vulnerabilities in transitive dependencies (viem, zod) — report upstream first; we'll respond to coordinated disclosures
• Vulnerabilities in the consumer's own bundler / RPC / signer infrastructure
• Theoretical timing attacks against Date.now() / setTimeout granularity
• Issues requiring physical access to the user's machine, signer, or HSM

The non-custodial path is zero-custody by construction:

• No private keys. The SDK never receives or persists the consumer's private key. Signing happens on the consumer's infrastructure (HSM, Fireblocks, web3signer, viem account, etc.).
• No relay. Signed UserOps go to the consumer-configured bundler, never to a Kash-hosted service.
• No paymaster. The smart account pays its own gas in ETH. The SDK does not sponsor gas or bundle paymaster data.
• No telemetry. The SDK does not phone home. There is no usage analytics, no error reporting to Kash, no opt-out required.

If you find a path that violates any of these guarantees, that's an in-scope vulnerability.

While the package is 0.x, only the latest published 0.x.y receives security fixes. After 1.0, we'll publish a long-term-support policy here.

Acknowledged researchers are listed in CHANGELOG.md against the patched release.