Security audit

webtest

Security checks across malware telemetry and agentic risk

Overview

This skill is a prompt-only website testing helper whose browser automation is disclosed and mostly scoped to user-authorized test targets, with some cautions around checkout flows and example credentials.

Install only if you intend to let the agent operate a browser against sites you are authorized to test. Use disposable test accounts, avoid production payments or real orders unless explicitly intended, and keep generated reports/screenshots local because they may reveal site data.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
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
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (5)

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The trigger list includes broad, common phrases such as '网页测试' and '测一下这个网站', which can match ordinary user conversation and cause the skill to activate outside clearly intended contexts. Because this skill can drive a browser and perform automated interactions, accidental invocation may lead to unintended navigation, testing actions, or handling of sensitive URLs and test credentials.

Vague Triggers

Medium

Confidence: 89% confidence
Finding: The activation description does not define sufficient boundaries for when the skill should run, beyond a loose set of keywords. In a browser-automation skill, missing context checks increases the chance of the agent acting on unrelated prompts, ambiguous references, or unauthorized targets, which is more dangerous than in a read-only informational skill.

Vague Triggers

Medium

Confidence: 88% confidence
Finding: The trigger phrases include very generic natural-language intents like "test this website" and Chinese equivalents, which can overlap with ordinary user requests that are not specifically meant to invoke this skill. This raises the risk of unintended activation, causing the agent to navigate to arbitrary URLs, perform browser actions, and potentially handle credentials or test flows without the user explicitly choosing this capability.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The example embeds plaintext test account credentials directly in documentation, which normalizes unsafe secret-handling and can lead users to reuse or expose real credentials in prompts, logs, screenshots, or reports. In an agent-driven browser testing skill, these values may be propagated across execution traces and artifacts, increasing the chance of credential disclosure beyond the immediate example.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The example demonstrates submitting an order without any warning or safeguard that this action may create a real transaction, charge an account, mutate backend state, or trigger downstream fulfillment. Because this skill is designed to execute browser actions from natural language, omission of a test-environment-only warning makes accidental production use more likely and increases the risk of unintended purchases or operational impact.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.