ClawHub
Back to skill

Security audit

Zeal Agent Wallet

Security checks across malware telemetry and agentic risk

Overview

This wallet skill is functional but needs Review because it gives an agent persistent transaction-proposal authority while hiding important Safe/delegate details and allowing proposals without confirmation.

Install only if you intentionally want an agent to create a persistent local wallet key and propose transactions for your Zeal Wallet. Require a clear approval step before every proposal, including recipient, value, calldata meaning, network, operation type, and purpose. Be especially cautious with DelegateCall, and confirm in Zeal/Safe how to revoke the agent’s proposer/delegate permission, not just delete local config.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Lp3

Medium
Category
MCP Least Privilege
Confidence
81% confidence
Finding
The skill directs the agent to run npm-based commands that perform networked transaction proposal operations, yet no corresponding permissions are declared. Undeclared network capability weakens security review and user/operator awareness, making it easier for a wallet-affecting skill to perform external actions without explicit policy scrutiny.

Tp4

High
Category
MCP Tool Poisoning
Confidence
89% confidence
Finding
The skill description understates and obscures sensitive behavior: it generates and stores a private key on disk, manipulates Safe-specific configuration while telling the agent to hide that terminology, and claims capabilities such as delegate wallet management that are not actually implemented. This mismatch can mislead users and reviewers about custody, persistence, and transaction-related risk, increasing the chance of unsafe consent and operational misuse.

Natural-Language Policy Violations

Medium
Confidence
85% confidence
Finding
The skill instructs the agent to conceal underlying Safe/Gnosis and delegate concepts from the user by forcing specific terminology. Mandated euphemistic wording reduces transparency about the real system being used and can prevent users from understanding security properties, researching risks, or giving informed consent for wallet configuration changes.

Missing User Warnings

High
Confidence
95% confidence
Finding
The skill explicitly authorizes proposing wallet transactions without user confirmation because owner approval happens later in the app. Even if execution still requires approval, unauthorized proposals can be used for social engineering, approval fatigue, transaction spam, and stealthy preparation of malicious calls such as dangerous DelegateCall payloads in a high-risk wallet context.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.exposed_secret_literal

File appears to expose a hardcoded API secret or token.

Critical
Code
suspicious.exposed_secret_literal
Location
src/setup.ts:37