Security audit

智能缓存工具-免费版

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a cache-management guide, but it asks for broad agent tools and contains unclear scope around triggering, disk persistence, and external API examples.

Review this skill before installing. It appears to be a cache helper rather than malware, but only use it for explicit cache-management tasks, avoid persisting secrets or private file/API contents unless you choose a safe path and permissions, and be careful with cache clear/delete operations. The author should narrow the trigger text and clarify network and disk-persistence behavior.

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 (6)

Intent-Code Divergence

Medium

Confidence: 95% confidence
Finding: The skill claims it is purely local and does not involve external API calls, yet its examples instruct calling remote APIs and fetching network data. This mismatch can mislead users and host agents into allowing network-capable behavior under a false local-only trust assumption, increasing the risk of unintended data egress or policy bypass.

Intent-Code Divergence

High

Confidence: 97% confidence
Finding: The free-edition limitation says only local in-memory caching is supported, but the document also advertises saving cache contents to disk. This contradiction matters because persistence can write sensitive cached material to the filesystem, undermining user expectations about data lifetime and storage location.

Intent-Code Divergence

Medium

Confidence: 91% confidence
Finding: The trigger-condition text describes analytics, reporting, and visualization scenarios unrelated to cache management. Overbroad and off-scope activation criteria can cause an agent to invoke this skill in inappropriate contexts, leading to unnecessary execution authority such as file access or command execution being applied to unrelated user requests.

Vague Triggers

High

Confidence: 96% confidence
Finding: The trigger condition is broad enough to match many unrelated requests, despite the skill having powerful tools like exec, grep, glob, and read. In an agent environment, this increases the chance of accidental or policy-bypassing invocation, where a cache tool is activated for generic data-analysis tasks and gains unnecessary access to local resources.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The skill supports deletion and full cache clearing but does not warn that these operations are destructive and may remove valuable cached state. In an agent setting, omission of such warnings can lead to accidental data loss or service degradation if a model invokes cleanup actions without informed user consent.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The persistence feature writes cached content to disk but does not warn that cached data may include sensitive API responses, file contents, or computed results. This can expose secrets or private data through local files, backups, shared directories, or weak file permissions.

Static analysis

No suspicious patterns detected.