Security audit

网络安全评估引擎免费版

Security checks for vulnerabilities and agentic risk

Overview

This skill is a local security-check helper, but it under-discloses its use of shell commands, file writes, dependency audit tools, and possible external callbacks.

Review before installing. Use it only on a project directory you intend to scan, expect shell commands and dependency audit tools to run, and do not provide callback URLs or allow file writes/hooks unless you explicitly want those side effects. Treat any secret-scan output as sensitive because it may reveal credentials or private code paths.

SkillSpector

Vulnerability Patterns

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
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (5)

Description-Behavior Mismatch

Medium

Confidence: 85% confidence
Finding: The skill markets itself as a pure knowledge-driven, no-extra-tools security checker, yet it instructs the agent to run shell commands and local auditing tools against the user's project. This mismatch can cause users or agents to grant broader execution privileges than expected, increasing the risk of unintended code execution, data exposure, or disruptive scans on local repositories.

Intent-Code Divergence

Medium

Confidence: 83% confidence
Finding: The documentation states that no extra tools or API keys are needed, but later depends on tools such as npm audit, pip-audit, grep, and references network/API behavior. This can mislead users about environmental requirements and side effects, causing unexpected network access or command failures in sensitive environments.

Description-Behavior Mismatch

Medium

Confidence: 81% confidence
Finding: The stated purpose is lightweight local self-checking, but the interface includes callback URLs and broader external integration behavior. Expanding from local review into outbound communication creates a larger attack surface, including the risk of exfiltrating assessment results or sensitive project metadata to remote endpoints.

Vague Triggers

Medium

Confidence: 76% confidence
Finding: A vague activation model increases the chance that the skill will run in situations the user did not intend, especially because it has Read, Write, Edit, and Bash capabilities. Poor trigger constraints are dangerous in agent contexts because they can lead to unexpected scanning, file modification, or command execution on arbitrary projects.

Missing User Warnings

Medium

Confidence: 87% confidence
Finding: The skill describes file writing, API calls, and command execution without prominent up-front disclosure of their operational impact. In an agent setting, this can cause users to unknowingly authorize state-changing actions or outbound requests, which may expose code, secrets, or system metadata.

Static analysis

No suspicious patterns detected.