Security audit

面试模拟工具

Security checks for vulnerabilities and agentic risk

Overview

The skill is an interview simulator, but it requests command and file-write authority and includes broad, inconsistent operational instructions that do not fit that purpose.

Review before installing. This does not show clear malicious behavior or an executable payload, but users should only enable it if they are comfortable with an interview-practice skill requesting command execution and write-capable tooling. Prefer a version that removes exec/write, network diagnostics, secret-handling language, and generic data-operation claims.

SkillSpector

Vulnerability Patterns

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
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration

Findings (5)

Description-Behavior Mismatch

Medium

Confidence: 93% confidence
Finding: The skill is presented as a Markdown-based interview simulator, but this section broadens scope into generic create/query/modify/delete and import/export/save/convert operations with parameterized execution semantics. That kind of capability inflation can cause an agent to treat unrelated user input as operational instructions and unnecessarily invoke tools or handle data beyond the stated interview purpose.

Context-Inappropriate Capability

Medium

Confidence: 95% confidence
Finding: The skill advertises command execution and ping-based network diagnostics despite being a Markdown interview tool with no clear operational need for shell or network activity. Unnecessary mention of exec-oriented and diagnostic behavior increases the chance that an agent will perform system or network actions from untrusted conversational context, expanding attack surface well beyond the stated function.

Intent-Code Divergence

Medium

Confidence: 96% confidence
Finding: The file contains contradictory claims: it says the tool is pure Markdown with no extra API key, yet also discusses command execution, HTTPS communication, environment-variable secret handling, and network testing. These inconsistencies are dangerous because they obscure the real trust boundary and can mislead an agent or operator into granting tools or handling secrets that the skill should not require.

Vague Triggers

Medium

Confidence: 88% confidence
Finding: The description includes broad activation guidance such as use for efficiency, automation, batch processing, and workflow optimization, which is far wider than interview simulation. Overbroad invocation text can cause the skill to be selected in unrelated contexts, where its tool permissions and vague operational claims may be applied inappropriately.

Vague Triggers

Medium

Confidence: 86% confidence
Finding: The capability coverage line lists vague scenario keywords rather than precise interview-related triggers, making invocation boundaries unreliable. In a tool-bearing skill, ambiguous routing language increases the risk of accidental activation and misuse in contexts where the agent may overinterpret user input as actionable operations.

Static analysis

No suspicious patterns detected.