Security audit

Gold Tracker

Security checks across malware telemetry and agentic risk

Overview

The skill is a coherent gold-price tracker, but it asks the installing agent to create persistent cron jobs that run and delete local records without clear opt-in or uninstall guidance.

Review before installing. Only enable the cron section if you want this skill to run unattended, fetch market data every 30 minutes, and periodically mutate or delete its local alert/archive files. Ask for the exact cron entries first and keep removal commands available.

SkillSpector

By NVIDIA

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Rogue AgentSelf-Modification, Session Persistence
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration

Findings (7)

Context-Inappropriate Capability

Medium

Confidence: 95% confidence
Finding: The skill explicitly instructs the installing agent to modify the user's crontab to establish recurring execution of local scripts. This creates persistent scheduled execution and expands the skill's privileges beyond one-time analysis, which is security-sensitive even if framed as operational convenience for a gold-tracking workflow.

Missing User Warnings

Medium

Confidence: 98% confidence
Finding: The skill directs automatic crontab installation without a clear user-facing warning that this is a persistent system change. Silent persistence reduces user awareness and can normalize unattended execution of code, making future misuse or accidental breakage harder to detect.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The maintenance instructions include cleanup and archive deletion behaviors that remove historical data on a schedule, but the skill does not clearly warn the user about data retention limits or irreversible deletion. This can lead to unintended loss of logs and alerts, especially when combined with automated cron execution.

Vague Triggers

Medium

Confidence: 84% confidence
Finding: The configuration explicitly encourages the agent to 'broadly search current market drivers' rather than limiting retrieval to vetted domains or a bounded source list. In an agentic workflow, this can expand the trust boundary to arbitrary web content, increasing exposure to prompt injection, misinformation, and malicious pages that could influence downstream analysis or actions.

Vague Triggers

Medium

Confidence: 88% confidence
Finding: The news search section is labeled 'flexible, non-restrictive' and tells the agent to adapt search terms based on current factors without defining boundaries on where it may search or what sources are acceptable. This makes the skill more dangerous because news content is an external, attacker-controllable input channel, and broad autonomous searching can lead the agent into poisoned or manipulative sources.

Session Persistence

Medium

Category: Rogue Agent
Content: **安装时智能体应执行**： 1. 将上述模板中的 `/path/to/gold-tracker` 替换为实际安装路径 2. 写入用户 crontab（`crontab -e` 或 `crontab < file`） 3. 验证 `crontab -l` 中已包含相关条目 4. 检查 `logs/cron.log` 是否有写入权限
Confidence: 96% confidence
Finding: crontab -e

Session Persistence

Medium

Category: Rogue Agent
Content: **安装时智能体应执行**： 1. 将上述模板中的 `/path/to/gold-tracker` 替换为实际安装路径 2. 写入用户 crontab（`crontab -e` 或 `crontab < file`） 3. 验证 `crontab -l` 中已包含相关条目 4. 检查 `logs/cron.log` 是否有写入权限 **注意事项**：
Confidence: 88% confidence
Finding: crontab -l

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.