ClawHub
Back to skill

Security audit

HyperGrok Hyperliquid Websocket

Security checks for vulnerabilities and agentic risk

Overview

This skill provides read-only Hyperliquid WebSocket monitoring guidance with disclosed local logging and no evidence of hidden, destructive, or credential-stealing behavior.

Install only if you are comfortable with local files containing trading activity such as fills, order updates, timestamps, and account identifiers when account subscriptions are enabled. Use restricted file permissions, rotate or delete old logs, and avoid logging account streams on shared machines unless that is intentional.

SkillSpector

Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (1)

Missing User Warnings

Medium
Confidence
89% confidence
Finding
The Python example persists every received WebSocket message to a local log file, including account-specific fills and order updates when `HYPERLIQUID_ACCOUNT_ADDRESS` is set. Even though the stream is read-only, these logs can expose sensitive trading activity, identifiers, and timing data to other local users, backups, or later forensic access, and the skill does not warn users that per-account data will be stored indefinitely.

Static analysis

No suspicious patterns detected.