Control Kite AI chain-wallet via Telegram using OpenClaw for wallet creation, balance checks, transfers, session keys, and spending limits without running a...

Do not install or use this skill until the author clarifies how signing keys are handled. Specific points to request/verify: 1) Where/how the wallet private key is provided and stored (the code expects a private key but SKILL.md/metadata do not declare this). 2) Who controls the private key used to create/send funds (giving the key to the agent equals giving custody of funds). 3) Why session keys are added with an all-functions selector and whether they can be scoped more narrowly. 4) Confirm the contract address and RPC endpoint are legitimate and intended. If you test this, use an empty test account and minimal funds on the testnet, run the code in an isolated/sandbox environment, and consider having the author publish a provenance/homepage and explicit install/auth instructions before trusting it with real funds.

Type: OpenClaw Skill Name: kite-agent-smart-wallet-v3 Version: 3.0.2 The skill is classified as suspicious due to its reliance on a high-privilege `KITE_WALLET_PRIVATE_KEY` (in `kite-wallet.js`) that directly controls a central agent's wallet. The `/kite send` command allows users to initiate transfers from this central wallet, and the `addSessionKey` function grants broad permissions to session keys on user-created smart wallets. While these are intended functionalities for a wallet agent, they introduce significant financial risk if the OpenClaw agent's input processing is vulnerable to prompt or command injection, potentially leading to unauthorized fund transfers or compromise of user wallets. There is no evidence of intentional malicious exfiltration or backdoor installation; the risk stems from the powerful capabilities exposed to user input.