Security audit

搜索引擎

Security checks for vulnerabilities and agentic risk

Overview

This skill appears to be a search helper, but it asks for broad local read and command-execution authority that is not clearly needed for search.

Install only if you are comfortable giving a search-oriented skill local read and command-execution capabilities. Avoid using it with secrets, confidential search terms, private files, or credentials unless the publisher narrows the tool permissions and clarifies privacy handling.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
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

Findings (5)

Description-Behavior Mismatch

High

Confidence: 95% confidence
Finding: The documented output schema and surrounding narrative describe compliance, quality, scoring, and security evaluation behavior rather than search-result aggregation. This kind of capability mismatch can cause an agent or user to trust the skill for unrelated decision-making tasks, creating scope confusion that may lead to unsafe execution paths, misuse of permissions, or fabricated outputs presented as authoritative analysis.

Description-Behavior Mismatch

High

Confidence: 97% confidence
Finding: The skill claims unrelated capabilities such as file parsing, batch file handling, API response parsing, command execution, and error repair despite being presented as a multi-search-engine integration. Inflated capability claims are dangerous because they normalize broader tool use, can trick an agent into invoking local or execution tools unnecessarily, and expand the apparent operational scope beyond what users would reasonably expect.

Context-Inappropriate Capability

Medium

Confidence: 92% confidence
Finding: The manifest requests read, exec, glob, and grep even though the described purpose is web search aggregation. Unnecessary local and execution permissions violate least privilege and become materially more dangerous in agent environments because they can enable unintended file access, local reconnaissance, or command execution if the skill is misused or later extended with unsafe instructions.

Vague Triggers

Medium

Confidence: 88% confidence
Finding: The skill description presents broad, loosely bounded capabilities across security tooling, compliance checking, decision support, and automation without defining when the skill should activate or what tasks are out of scope. In an agent setting, ambiguous trigger boundaries increase the chance of over-invocation, tool misuse, and users unintentionally sending unrelated or sensitive tasks through a skill with broader permissions than expected.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The skill describes sending user queries to external search engines but does not clearly disclose that those queries may be transmitted to third parties. This is a meaningful privacy and compliance risk because users may submit sensitive prompts, internal identifiers, or regulated data under the mistaken assumption that the skill operates locally or privately.

Static analysis

No suspicious patterns detected.