Security audit
Security checks for vulnerabilities and agentic risk
This is a low-impact documentation workflow skill with overly broad activation wording but no executable code, credential handling, persistence, or destructive behavior.
Installers should know this skill may activate too easily for ordinary search or workflow requests. It appears safe as a documentation helper, but its trigger terms should be narrowed if precise routing matters.
The trigger phrases are broad enough to match generic terms like "multi", "search", and "engine", which can cause the skill to activate in unrelated contexts. In an agent ecosystem, that increases the chance of unintended invocation, context hijacking, or the wrong workflow being applied to user requests, which can degrade safety and reliability.
The manifest description says to use the skill when a user asks for broad terms like "work-productivity," "multi," "search," or needs any practical workflow or analysis support. That invocation scope is so broad that many unrelated prompts could match, increasing unintended routing and making the skill dangerous in orchestration environments where descriptions influence automatic selection.
The trigger keyword list includes extremely generic terms such as "multi", "search", "engine", "integration", and "global", which are common in unrelated user requests. This makes accidental invocation likely, causing the skill to activate outside its intended scope and potentially override or interfere with more appropriate skills or user workflows.
触发关键词包含“multi”“search”“engine”“global”“supports”等高度通用词，容易在与该技能无关的普通对话中被意外匹配并调用。误触发会让代理偏离用户真实意图，注入不相关流程或产物，在多技能环境中还可能压制更合适的技能选择。
The trigger sentence is overly broad and built from common everyday phrasing, which can cause the skill to activate in conversations where the user did not explicitly intend to invoke it. In an agent environment, unintended activation can route user requests into the wrong workflow, causing irrelevant actions, confusion, or accidental execution of automation steps under the wrong context.
The trigger sentences are generic and overlap heavily with ordinary user requests such as asking for help, needing a workflow, or invoking broad search-related tasks. In an agent ecosystem, this can cause unintended activation of the skill in contexts where the user did not specifically intend to use it, which may route tasks to the wrong workflow and produce unsafe, irrelevant, or misleading outputs.
This file is presented as a Chinese README, but key user-facing sections including the title, demand description, workflow description, keywords, and trigger phrases are written largely in English. That can effectively impose a language/locale on users without opt-in or justification, which conflicts with language-choice expectations.
The example trigger sentences are malformed and largely repeat marketing text instead of showing precise activation patterns. Poorly specified examples can train or bias routing systems toward noisy matches, compounding the already broad trigger surface and making unintended activation more frequent.
技能描述将适用范围定义为包含 work-productivity、multi、search、engine 等宽泛概念，边界不清晰，导致路由器或代理在大量模糊场景下都可能选择该技能。这种过宽选择面会扩大误用概率，使输出与用户需求错配，并可能干扰其他更精确、更安全的技能。
示例触发句几乎只是重复需求文案，没有给出判定边界、必备上下文或反例，容易让实现者误以为只要出现相关词就应调用该技能。虽然这更像设计疏漏而非恶意，但它会进一步放大前述宽泛触发逻辑带来的误调用风险。
The default prompt is phrased as a broad, natural-language trigger that can match common user requests about productivity, search, workflows, or implementation help without strong scoping constraints. Combined with allow_implicit_invocation: true, this increases the chance of unintended automatic invocation, which can route unrelated conversations into this skill and expand the attack surface for prompt injection, data overexposure, or unsafe tool use.
The activation guidance lacks precise boundaries and includes ambiguous examples, making it easier for unrelated requests to match the skill. Because this skill is framed as a general helper for workflows, checklists, analysis, and implementation support, ambiguous routing increases the chance of misclassification and unintended use across a wide range of normal productivity conversations.
