Security audit

Local LLM Setup Advisor

Security checks for vulnerabilities and agentic risk

Overview

This skill is a local-LLM setup guide with no executable code, though its activation wording is broader than ideal.

Install only if you want general help with local LLM setup and hardware-constrained AI workflows. Be aware it may activate on broad terms such as privacy or software-and-data, so review whether it is the right skill when working on unrelated tasks.

SkillSpector

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (9)

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The trigger sentences are generic and partially templated, which can cause the skill to activate for loosely related requests rather than clear local-LLM setup tasks. Over-broad activation increases the chance of inappropriate skill routing, causing the agent to inject irrelevant guidance, override better-matched skills, or respond in contexts where hardware/software advice was not requested.

Vague Triggers

Medium

Confidence: 96% confidence
Finding: The activation section lists phrases like "Help me" and "I need a practical workflow" wrapped around a long requirement sentence, which overlaps with common conversational language and does not clearly define when the skill should or should not activate. The file also provides no negative examples or explicit scope constraints to prevent unintended invocation.

Vague Triggers

Medium

Confidence: 97% confidence
Finding: The skill description is broad enough to match many general AI or software-help requests, which can cause the skill to activate outside its intended niche. Over-broad activation increases the chance of routing users into irrelevant or lower-safety guidance, and in a multi-skill system can override more appropriate specialized skills.

Vague Triggers

Medium

Confidence: 96% confidence
Finding: The trigger keywords contain vague, high-collision terms like 'privacy' and 'software-and-data' that are likely to appear in unrelated conversations. This can lead to accidental invocation, poor task routing, and increased attack surface if the skill is pulled into contexts it was not designed to handle.

Vague Triggers

Medium

Confidence: 97% confidence
Finding: The trigger keywords include broad everyday terms such as "privacy", which can match many unrelated user requests and cause the skill to activate unexpectedly. Overbroad activation is dangerous because it can route users into guidance they did not request, increasing the chance of irrelevant, misleading, or lower-priority security advice being injected into unrelated conversations.

Vague Triggers

Medium

Confidence: 89% confidence
Finding: The default prompt is broad and generic enough to match common help-seeking language, which can cause the skill to activate outside its intended scope. In a security-sensitive agent ecosystem, overbroad routing increases the chance of unintended data exposure, irrelevant tool use, or steering users into this skill when a more appropriate or safer skill should handle the request.

Vague Triggers

Medium

Confidence: 94% confidence
Finding: Enabling implicit invocation without tightly defined activation conditions allows the platform to call this skill based on loose semantic similarity rather than deliberate selection. Because this skill concerns broad software-and-data guidance, that can lead to accidental invocation on adjacent requests, increasing misrouting risk and unintended handling of user context.

Vague Triggers

Medium

Confidence: 98% confidence
Finding: The trigger sentence is so broad and templated that it can match ordinary user requests unrelated to a narrowly scoped skill invocation. This can cause the agent to activate the skill unexpectedly, increasing the chance of misrouting, prompt-surface expansion, and unintended influence over responses in benign conversations about local AI setups.

Natural-Language Policy Violations

Low

Confidence: 77% confidence
Finding: The zh-CN README contains a full English requirement statement in the main user-facing content, which can impose a language expectation inconsistent with the localized document. There is no indication that users may choose their preferred language or that the mixed-language content is intentional.

Static analysis

No suspicious patterns detected.