Security audit

JS SDK工具专业版

Security checks across malware telemetry and agentic risk

Overview

This is a documentation-style JavaScript AI SDK helper whose capabilities fit its purpose, though users should be careful with examples that send files, prompts, or webhook data to external services.

Install only if you want a JavaScript AI SDK integration guide. Before reusing the examples, review what data is sent to model providers, proxy servers, webhooks, external URLs, memory features, or code-execution tools, and keep API keys in environment variables rather than client code.

SkillSpector

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: 93% confidence
Finding: The trigger condition is defined very broadly as any scenario involving AI model calls, chat, agent orchestration, or LLM apps, without concrete exclusions or routing boundaries. In an agent ecosystem, this can cause the skill to activate for loosely related prompts and steer execution toward high-capability tooling, increasing the chance of unintended use of exec-enabled workflows or inappropriate handling of sensitive tasks.

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The capability coverage section enumerates an extremely broad set of keywords and claims support for all matching scenarios, creating overlap with many unrelated requests. Because the skill also declares powerful tools like exec, broad matching raises the risk of accidental invocation and over-privileged handling in contexts the skill was not specifically designed for.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The examples encourage sending local files, base64 content, browser File objects, and loading external skill content by URL, but do not prominently warn that this may transmit sensitive data to remote AI services or third-party hosts. In an enterprise-oriented SDK skill, users may copy these patterns into production workflows and unknowingly exfiltrate confidential files, prompts, or internal documentation.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The documentation demonstrates a webhook tool that posts to an external Slack endpoint and uses a secret, but it does not warn that tool inputs may be transmitted to a third party. In an agent SDK context, developers may pass user content, model outputs, or sensitive operational data into this webhook, creating an unintentional data exfiltration/privacy risk.

Missing User Warnings

Medium

Confidence: 86% confidence
Finding: The proxy examples show routing SDK requests through server endpoints using an API key, but they do not disclose that user prompts, chat history, and tool-related data may transit and be processed by the application server. This can mislead implementers into deploying the proxy without appropriate privacy notice, logging controls, authentication, or server-side access restrictions.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.