Back to skill

Security audit

HyperGrok Hyperliquid Setup

Security checks for vulnerabilities and agentic risk

Overview

The scanners mark this skill suspicious because it can store a trading private key in a plaintext file and describes setup steps as read-only even when they install software or change files; scan coverage is partial.

Consider avoiding installation until the private-key file fallback is removed or you can confirm a secure secret provider is used, and until installation and file changes are clearly disclosed and authorized. The scan was partial: although all three listed components were scanned, some local references could not be resolved, so the evidence is incomplete.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
88% confidence
Finding

The advertised secure-secret-store provisioning has an explicit fallback that persists the trading private key in a plaintext local file. Although user-controlled creation and restrictive permissions reduce exposure, the skill explicitly describes a computer shared by Bots; those permissions do not isolate processes running as the same user, and compromise of a trading key can cause financial losses without withdrawal access.

Content

No source excerpt is available for this finding.

Sections described as read-only include SDK installation and persistent filesystem changes.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The blanket statement that sections 1–3 are read-only and safe at any time incorrectly includes dependency installation and persistent filesystem and package-configuration changes. An agent relying on that statement could perform software installation, including execution of package-supplied code, under authorization intended only for inspection; the commands themselves are otherwise consistent with environment setup.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 61)May include surrounding context.

bash
for base in https://api.hyperliquid.xyz https://api.hyperliquid-testnet.xyz; do
  printf '%s -> ' "$base"
  curl -sS -m 10 -X POST "$base/info" -H 'Content-Type: application/json' \
    -d '{"type":"allMids"}' | jq -r 'if type=="object" then "ok, \(length) mids, BTC \(.BTC)" else "unexpected: \(.)" end'
done

The key loader permits plaintext-file provisioning despite the manifest's secure-secret-store boundary.

Medium
Category
Excessive Agency
Confidence
97% confidence
Finding

The loader silently falls back from an environment-provided secret to a persistent plaintext key file, weakening the stated secure-store provisioning model. This duplicates the underlying issue in TP4: mode 600 protects against other ordinary OS users, but not other Bots sharing the owner's identity, and the fallback can also retain a stale credential after secret-store updates.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · SKILL.md (reported line 93)May include surrounding context.

If the desk's Grok Bot setup exposes secrets to the computer as environment variables, scripts read HYPERLIQUID_PRIVATE_KEY from the environment. If it does not, the fallback is a file the user creates while in control of the computer:

bash
mkdir -p ~/.hyperliquid && chmod 700 ~/.hyperliquid
# the user pastes the key into this file themselves, then:
chmod 600 ~/.hyperliquid/api-wallet.key