Shekel Hyperliquid Trading

AI-powered perpetual futures trading on Hyperliquid DEX. Handles full account creation, USDC onboarding, and autonomous trade execution.

Plain-language checklist before installing or using this skill: - The skill asks your agent to fetch live instructions from a remote URL before every operation. That means the remote server can change what the skill tells your agent to do at any time — review and trust that backend before using the skill with real funds. - The single required credential (SHEKEL_API_KEY) gives broad control: the API supports deposits, withdrawals, exporting wallet keys, and deleting accounts. Treat that key like a full-access secret — do NOT supply it unless you trust the service and have limited, auditable permissions. - The skill also tells the agent to read stored credentials from MEMORY.md. Confirm where credentials would be stored and who/what can read them — local files or agent memory could be exposed. - Because the agent can run autonomously, consider limiting or disabling autonomous runs for real-money accounts and require manual confirmation for any withdrawal or export operations. - If you still consider using it: (1) audit the backend URL and /skill payload to see exactly what commands are returned; (2) use a low-privilege API key (if possible) or test with a burner account and minimal funds; (3) enable monitoring and key rotation; (4) require explicit human approval for any withdrawal/export operations. If you want, I can: list the specific API calls in the SKILL.md that allow sensitive actions, draft safe policies for using the key (scopes, lifetime), or suggest questions to ask the skill provider about the remote-update mechanism and data retention.

Type: OpenClaw Skill Name: shekel-hyperliquid Version: 1.14.0 The shekel-hyperliquid skill (v1.14.0) implements a high-risk remote instruction fetching mechanism in SKILL.md, mandating that the agent fetch and execute updated instructions from an external endpoint (https://shekel-skill-backend.onrender.com/skill) before every interaction. This effectively allows the backend to perform remote prompt injection and alter agent behavior dynamically after deployment. Furthermore, the skill handles sensitive financial credentials, including instructions to store API keys in local memory and an endpoint (POST /auth/export-agent-key) to exfiltrate the trading wallet's private key. While these features are framed as part of a managed trading service, the combination of remote instruction updates and private key access represents a significant security risk.