Security audit

游戏资产生成-专业版

Security checks across malware telemetry and agentic risk

Overview

This skill is mostly a Markdown game-asset helper, but it requests broad read/write/command authority with loose triggers and underdefined network, file, and execution safeguards.

Install only if you are comfortable with a game-asset skill that can read and write files and may run shell commands. Keep it limited to a dedicated project workspace, review any generated command before execution, avoid using callback_url unless you control the endpoint, and do not provide secrets unless the agent explicitly explains why they are needed.

SkillSpector

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
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands

Findings (6)

Intent-Code Divergence

Medium

Confidence: 91% confidence
Finding: The documentation makes conflicting security and credential-handling claims: it says no extra API key is needed while also instructing users to manage API keys via environment variables. This inconsistency can cause operators to misconfigure the skill, underestimate secret exposure risk, or assume hidden external integrations are harmless when they may still process credentials or sensitive data.

Intent-Code Divergence

Medium

Confidence: 95% confidence
Finding: The file claims commands run in a 'safe sandbox' but elsewhere grants generic exec capability without defining actual enforcement boundaries, allowed commands, or input validation rules. That mismatch can lead users to trust unbounded command execution and pass attacker-controlled inputs into shell commands, increasing the risk of arbitrary command execution or system misuse.

Vague Triggers

Medium

Confidence: 88% confidence
Finding: The activation text is overly broad and tied to generic productivity phrases such as efficiency improvement, automation, batch processing, and workflow optimization. Overbroad triggers can cause the skill to activate in unrelated contexts, exposing read/write/exec capabilities more often than necessary and increasing the chance of unintended high-privilege actions.

Vague Triggers

High

Confidence: 97% confidence
Finding: The keyword coverage section includes extremely vague trigger tokens such as 'Use', 'when', and broad workflow language without scope constraints. In a skill with exec, read, and write tools, such unconstrained activation greatly increases accidental invocation and expands the attack surface for tool misuse in unrelated conversations.

Missing User Warnings

Medium

Confidence: 84% confidence
Finding: The skill describes create/query/export operations and includes write capability, but it does not clearly warn users that local files may be created or modified. Without explicit disclosure and consent, users may unknowingly permit file writes, leading to unintended overwrites, data loss, or persistence of generated content in sensitive directories.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The documented callback_url parameter enables sending data to remote endpoints but provides no privacy or security warning about external transmission. This can result in users unintentionally exfiltrating prompts, generated assets, metadata, or project details to third-party servers they do not control or fully trust.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.