Security audit

gamegeeking Community Engagement

Security checks across malware telemetry and agentic risk

Overview

This skill openly automates GameGeeking account use and public posting, but it also handles passwords, can edit public profiles, and tells the agent to make automated activity look like an ordinary player.

Review before installing. Use only a dedicated GameGeeking account, do not share a primary password, require explicit approval for the exact post/comment and every profile edit, and avoid using the skill to disguise automated, sponsored, or coordinated activity as ordinary user participation.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (6)

Context-Inappropriate Capability

Medium

Confidence: 96% confidence
Finding: The skill goes beyond posting/commenting and explicitly instructs the agent to groom profile attributes such as nickname, bio, and avatar to make a newly created account appear more natural. In context, this is not incidental UX polish: it is framed as a way to reduce suspicion and improve the credibility of automated activity, which materially facilitates deceptive impersonation and covert influence operations.

Vague Triggers

Medium

Confidence: 78% confidence
Finding: The trigger phrases include broad language such as '去社区聊聊' that can match ordinary conversation requests and cause the skill to activate for high-impact actions like logging in, registering, and posting. Because this skill performs external side effects on a live site, ambiguous activation increases the risk of unintended account actions and unauthorized posting.

Missing User Warnings

High

Confidence: 95% confidence
Finding: The skill instructs the agent to solicit and use usernames and passwords for a third-party service without any warning about sensitive credential handling, storage, masking, or safer alternatives. This creates a direct pathway for credential exposure, insecure retention in logs, and misuse of user secrets by the agent or surrounding infrastructure.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: From the outset, the skill is designed to perform live registration, login, posting, commenting, and optional profile modification on a user-facing platform, but it does not present a clear up-front warning that it will create or alter public account and community data. Users may invoke it without appreciating that it causes irreversible external side effects visible to others.

Ssd 4

High

Confidence: 98% confidence
Finding: This section explicitly instructs the agent to make a fresh account look less empty, less temporary, and more like a real player before posting. That is a classic trust-building and anti-detection tactic designed to conceal the automated or orchestrated nature of the account, increasing the effectiveness of deceptive engagement and reducing the chance of moderation scrutiny.

Ssd 4

High

Confidence: 99% confidence
Finding: The guidance repeatedly tells the agent to sound like a 'real player,' avoid looking like an operator or AI, avoid revealing automated origin, and tailor content to blend into the community. In this context, those instructions are not harmless style advice; they operationalize stealthy impersonation and covert content seeding on a third-party community.

VirusTotal

61/61 vendors flagged this skill as clean.

View on VirusTotal