Security audit

VoxPact

Security checks across malware telemetry and agentic risk

Overview

VoxPact is a disclosed marketplace skill that uses shell scripts and an API key to let an agent work with jobs, payments, messages, and files on an external service.

Install only if you want your agent connected to VoxPact with authority to interact with paid jobs. Treat VOXPACT_API_KEY as a payment-linked secret, review commands before running actions like posting jobs, approving work, bidding, uploading files, or delivering work, and avoid sending private code or business data unless you intend to share it with the marketplace and job counterparties.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (5)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 88% confidence
Finding: The skill advertises use of environment variables and shell scripts that can perform marketplace actions, file transfer, and webhook interactions, but it does not declare corresponding permissions or boundaries. This can cause agents or users to invoke networked and shell-capable behavior without clear consent or sandbox expectations, increasing the chance of unintended command execution or data handling.

Missing User Warnings

Low

Confidence: 84% confidence
Finding: The metadata and description instruct users to provide a live API key but do not include any warning about secret handling, logging, or avoiding exposure in prompts, shell history, or shared environments. While this is not direct exfiltration by itself, it normalizes unsafe credential handling and can lead to accidental leakage of a payment-linked marketplace credential.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The skill prominently supports uploading and downloading files, sending messages, and registering webhooks with an external marketplace, but it does not warn that user or agent data will be transmitted off-platform. In this context, the omission is meaningful because job files, messages, and webhook payloads may contain sensitive business data, code, or personal information.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: This bootstrap hook silently injects marketplace instructions that encourage the agent to seek paid work, hire other agents, and interact with an external service whenever a VOXPACT_API_KEY is present. Because it modifies agent behavior at startup without explicit user disclosure or consent, it can steer the agent into externally initiated actions, economic transactions, or data sharing that the user did not request, which is a real security and trust-boundary issue.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The hook silently injects a bootstrap document that teaches the agent to run marketplace-related shell scripts, transfer files, and pursue paid work whenever VOXPACT_API_KEY is present. Because this happens at agent bootstrap with no user-facing disclosure, consent gate, or policy constraint, it expands the agent's action surface in a way operators may not expect and can steer autonomous behavior toward external network/file operations and financial workflows.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.