Security audit

DNS网络诊断基础版

Security checks for vulnerabilities and agentic risk

Overview

This is a plausible network-diagnostics skill, but it under-discloses live outbound probes and privileged local network changes that an agent may run automatically.

Install only if you are comfortable with an agent running local shell commands for network troubleshooting. Treat every DNS, curl, nc, ping, and openssl example as live traffic to the named host or resolver, and require confirmation before sudo commands, package installs, cache flushes, /etc/hosts edits, or repeated API timing requests.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection

Findings (11)

Intent-Code Divergence

Medium

Confidence: 97% confidence
Finding: The skill states that all diagnostic commands run locally and will not send data to external services, but many documented commands explicitly contact remote DNS resolvers and remote HTTP/TLS endpoints. This is dangerous because users may disclose internal hostnames, target domains, or connectivity metadata under a false privacy assumption.

Vague Triggers

Medium

Confidence: 90% confidence
Finding: The trigger condition says the skill should be used for broad categories like code generation, programming assistance, debugging, and deployment, which exceed the stated networking-diagnostics scope. Overbroad triggers can cause the agent to invoke live network and system commands in unrelated contexts, increasing the chance of unintended execution.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The skill introduces commands that perform live DNS, HTTP, TLS, and port connectivity checks without clearly warning that target information will be transmitted to external hosts. In a tool with exec capability, the lack of an upfront network activity warning can lead users to unknowingly initiate outbound probes.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: This section documents local system-modifying commands such as flushing DNS caches and editing hosts-related configuration areas without a clear warning about operational impact. Such actions can alter resolution behavior, disrupt applications, or require elevated privileges, especially when an agent may execute commands automatically.

External Transmission

Medium

Category: Data Exfiltration
Content: 总耗时: %{time_total}s HTTP状态码: %{http_code} 响应大小: %{size_download} bytes " https://api.example.com/endpoint # ... # 查看响应头 curl -sI https://api.example.com/endpoint
Confidence: 93% confidence
Finding: The documented curl timing command sends a live HTTP request to an external endpoint, transmitting target and client connectivity metadata outside the local machine. In the context of an exec-enabled skill, this is security-relevant because it performs outbound communication and may expose internal troubleshooting targets or environment details.

External Transmission

Medium

Category: Data Exfiltration
Content: " https://api.example.com/endpoint # ... # 查看响应头 curl -sI https://api.example.com/endpoint # ... # 跟随重定向 curl -sIL https://example.com
Confidence: 93% confidence
Finding: The header-fetch and redirect-follow commands contact external hosts directly, causing network transmission to third-party services. Without strong consent and disclosure, users may unintentionally leak which services they are troubleshooting or trigger requests to untrusted destinations.

External Transmission

Medium

Category: Data Exfiltration
Content: 服务器处理: %{time_starttransfer}s 总耗时: %{time_total}s 下载大小: %{size_download} bytes " https://api.example.com/endpoint # ... # 判断瓶颈位置 TOTAL=$(curl -o /dev/null -s -w "%{time_total}" https://api.example.com/endpoint)
Confidence: 93% confidence
Finding: The API latency analysis example repeatedly issues live curl requests to an external endpoint, increasing both data exposure and the operational footprint of the action. Repeated outbound requests can also be problematic if aimed at sensitive internal APIs or unintended third-party services.

Sudo/Root Execution

Medium

Category: Privilege Escalation
Content: cat /etc/hosts # 所有平台 # ... # 刷新DNS缓存 sudo dscacheutil -flushcache; sudo killall -HUP mDNSResponder # macOS sudo systemd-resolve --flush-caches # Linux(systemd) ipconfig /flushdns # Windows # ...
Confidence: 95% confidence
Finding: This finding points to another sudo invocation on the same cache-flush line, again indicating elevated system modification commands embedded in normal diagnostic guidance. The danger is amplified by the skill's use of exec and the absence of clear execution boundaries between observation and modification.

Sudo/Root Execution

Medium

Category: Privilege Escalation
Content: cat /etc/hosts # 所有平台 # ... # 刷新DNS缓存 sudo dscacheutil -flushcache; sudo killall -HUP mDNSResponder # macOS sudo systemd-resolve --flush-caches # Linux(systemd) ipconfig /flushdns # Windows # ...
Confidence: 95% confidence
Finding: This finding points to another sudo invocation on the same cache-flush line, again indicating elevated system modification commands embedded in normal diagnostic guidance. The danger is amplified by the skill's use of exec and the absence of clear execution boundaries between observation and modification.

Sudo/Root Execution

Medium

Category: Privilege Escalation
Content: # ... # 刷新DNS缓存 sudo dscacheutil -flushcache; sudo killall -HUP mDNSResponder # macOS sudo systemd-resolve --flush-caches # Linux(systemd) ipconfig /flushdns # Windows # ... # 查看systemd-resolved状态
Confidence: 94% confidence
Finding: The Linux cache-flush example also requires sudo and changes system resolver state. In an automated agent workflow, privileged commands that alter networking behavior can impact the host beyond the current troubleshooting session.

Chaining Abuse

High

Category: Tool Misuse
Content: cat /etc/hosts # 所有平台 # ... # 刷新DNS缓存 sudo dscacheutil -flushcache; sudo killall -HUP mDNSResponder # macOS sudo systemd-resolve --flush-caches # Linux(systemd) ipconfig /flushdns # Windows # ...
Confidence: 88% confidence
Finding: The use of command chaining with multiple sudo operations on a single line increases the blast radius of accidental execution and makes it easier for an agent to perform several privileged actions in one step. In an exec-enabled skill, chained privileged commands reduce opportunities for user review between actions.

Static analysis

No suspicious patterns detected.