Security audit
Security checks for vulnerabilities and agentic risk
This skill is a local LLM setup advisor with no executable payload, persistence, credential handling, or destructive behavior, though its activation wording is broader than ideal.
This skill is reasonable to install for local LLM setup help. Be aware it may activate on broad privacy or workflow requests, so invoke it explicitly when you want local inference guidance and prefer clearer trigger wording in future versions.
The trigger phrases are broad and natural-language-like, which increases the chance the skill is invoked unintentionally for loosely related requests. In an agent ecosystem, ambiguous activation can route user prompts into an unintended workflow, causing mis-scoping, confusing outputs, or accidental disclosure of user context to the wrong skill.
The trigger sentences are broad, repetitive, and partially truncated, which can cause the skill to activate outside the user's clear intent. In an agent environment, overbroad activation can route unrelated prompts into this skill, producing inappropriate guidance, bypassing better-matched skills, or increasing prompt-injection exposure by invoking the wrong workflow.
This file is labeled as a Chinese README, but key user-facing content such as the demand description and target users is written in English. That creates an implicit language/locale inconsistency without giving the user an opt-in choice or explaining that mixed-language output is intended.
The skill description and activation guidance are broad enough that the skill may trigger for loosely related requests, causing incorrect routing or unintended invocation. In an agent environment, overly permissive triggers can let a non-specialized skill intercept user tasks and produce misleading advice, reducing reliability and potentially bypassing more appropriate safeguards or domain-specific handling.
The example trigger phrases use everyday-language patterns like 'Help me' and 'I need a practical workflow,' which are common across many unrelated tasks and can cause accidental activation. This increases the chance of prompt-routing errors, where the skill is selected based on vague phrasing rather than true relevance, degrading system behavior and confusing users.
The usage condition in the skill description is broad and blends several generic request types like workflows, checklists, analysis, and implementation support without strong scoping constraints. In a multi-skill environment, this can lead to unintended invocation for loosely related requests, increasing the risk of misrouting, policy bypass through skill confusion, or inappropriate reliance on this skill for topics outside local LLM setup.
The trigger keyword list includes the very broad term "privacy", which can overlap with many unrelated user requests and cause the skill to activate outside its intended scope. Over-broad activation increases the chance of incorrect routing, unnecessary exposure of this skill's instructions, and degraded security boundaries between skills, even though it does not by itself execute harmful actions.
The skill enables implicit invocation and uses broad activation text tied to generic topics like local LLMs, software, and workflows, without clear boundaries on when it should trigger. This can cause the agent to invoke the skill in situations the user did not clearly request, increasing the chance of prompt-scope confusion, unintended tool influence, or interception of adjacent requests.
The invocation guidance lacks clear boundaries and includes broad keywords and trigger phrases that could overlap with many normal requests about AI, software, or privacy. This increases accidental invocation and misrouting risk, especially in agent systems where skill selection influences what instructions, assumptions, and tools are applied to the user's request.
The trigger sentence is so generic that it can match ordinary user phrasing and cause the skill to activate outside its intended scope. Over-broad routing can send unrelated conversations into this skill, leading to incorrect assistance, reduced user control, and possible policy bypass if downstream behavior assumes the trigger was high-confidence.
The natural-language file descriptions declare English and Chinese variants, but do not state how the user's preferred language is selected or whether language choice is optional. This can create a locale-policy concern if the skill defaults to a language without explicit user opt-in.
The example trigger phrase "I need a practical workflow for..." is highly generic and resembles ordinary help-seeking language, which can bias matching toward accidental activation. While the surrounding topic text reduces severity somewhat, generic examples still weaken trigger precision and make the skill easier to invoke unintentionally.
No suspicious patterns detected.
