Near Subaccount

Create, list, delete, and manage NEAR subaccounts with bulk distribution operations.

Before installing or enabling this skill: 1) Inspect scripts/subaccount.js and package.json to confirm what the code does (look for file reads, network endpoints, child_process usage, or attempts to read ~/.near-credentials). 2) Confirm how the near-subaccount CLI is installed — the SKILL.md documents a command but provides no install steps. 3) Be aware this will need NEAR CLI credentials and access to the master account (private keys); do not run it on a machine that holds keys you cannot risk exposing. 4) If you plan to allow autonomous model invocation, avoid doing so until you verify the code does not exfiltrate credentials or perform unexpected network calls. 5) If anything is unclear, request the skill author supply installation instructions, a security/privacy note about credential use, and a code audit or reproducible build steps before proceeding.

Type: OpenClaw Skill Name: near-subaccount Version: 1.0.1 The `scripts/subaccount.js` file uses `child_process.exec` to run NEAR CLI commands and `fs.promises.readFile` to read a JSON file. While these capabilities are necessary for the skill's stated purpose, the script does not perform explicit input sanitization on arguments passed to `execAsync` (e.g., `subaccountName`, `masterAccount`, `accountId`), making it vulnerable to command injection if malicious input is provided by the agent or user. Furthermore, the `bulkDistribute` function allows reading an arbitrary file path (`filePath`) specified by the user, which could be exploited for data exfiltration of sensitive local files (e.g., `~/.ssh/id_rsa`, `/etc/passwd`) if the agent is prompted to do so. There is no clear evidence of intentional malicious behavior, but the high-risk capabilities and lack of input sanitization make it suspicious.