Security audit
Security checks for vulnerabilities and agentic risk
This is a documentation-only workflow helper with broad activation wording but no hidden code, persistence, credential access, or destructive behavior.
Before installing, be aware that this skill may be invoked more often than intended because its trigger language is broad and implicit invocation is enabled. It appears safe as a documentation-only workflow helper, but users who want precise routing should narrow or disable the generic triggers.
The use-condition says to use the skill whenever a user asks for broad categories like work-productivity or practical workflow support, which are expansive and underspecified. This ambiguity can make the skill match many benign requests unrelated to self-improving agent workflows, leading to misrouting, unwanted instruction injection, and reduced reliability of the agent system.
The trigger list includes extremely generic terms like "self," "improving," and "proactive," which are common in ordinary prompts and can cause the skill to activate far outside its intended scope. Over-broad activation increases the chance the agent injects irrelevant workflow guidance into unrelated tasks, creating prompt-routing confusion and potentially displacing more appropriate skills or safety policies.
触发关键词包含极其宽泛的通用词，如“self”“improving”“proactive”，会让技能在大量无关场景下被误触发。误触发会扩大该技能对会话的影响范围，导致错误路由、无关指令注入到当前任务，甚至让后续更敏感的技能选择逻辑被干扰。
The trigger sentences are generic enough to match many ordinary user requests involving help, practical workflows, or productivity, which can cause the skill to activate outside its intended scope. Overbroad activation increases the chance that this skill intercepts unrelated tasks, leading to misrouting, unintended behavior, or abuse of broad delegation pathways in agentic environments.
The trigger sentences are extremely generic ('Help me...', 'I need a practical workflow...') and map to broad everyday productivity requests, which can cause the skill to activate outside its intended scope. In an agent ecosystem, overbroad activation can hijack unrelated tasks, suppress more appropriate skills, or route user input into a self-improving/proactive workflow that may take unintended actions or produce irrelevant artifacts.
The trigger phrases are broad enough to activate on generic terms like 'self', 'improving', and 'proactive', which can cause the skill to be invoked in contexts far outside its intended scope. In an agent framework, this increases the risk of unintended routing, confusing workflow selection, and accidental execution of capability-bearing instructions when the user did not explicitly request this skill.
The example trigger phrases are nearly tautological and provide little boundary-setting, so they do not help distinguish valid invocations from ordinary discussion. Weak examples reinforce over-triggering behavior and make downstream routing decisions less predictable, especially in systems that rely on examples as activation hints.
技能描述把适用范围定义为包含 work-productivity、self、improving、proactive 或“需要实用流程/分析支持”的请求，边界非常模糊，几乎可覆盖大量普通生产力和改进类任务。这会导致技能被过度匹配，在不适合的上下文中介入，降低系统对用户意图的精确识别，并增加被滥用为通用入口的风险。
示例触发句几乎可被任何相似表述套用，未体现调用该技能所需的明确边界、输入要求或任务约束。这会强化模糊匹配行为，使编排器或用户倾向于在宽泛需求下调用该技能，增加误用和上下文污染的概率。
The default prompt and invocation description use broad, common-language phrases such as 'help me' and generic productivity terms that can match many unrelated user requests. In combination with agent routing, this can cause unintended invocation of the skill in contexts the user did not explicitly request, exposing users to unexpected behavior or prompt influence from this skill.
Enabling implicit invocation without tightly scoped trigger constraints increases the chance that this skill activates automatically on broad productivity-related requests. Because the skill is framed around proactive and self-improving workflows, accidental activation could let it influence conversations or actions beyond the user’s intent, making prompt-routing abuse and scope creep more likely.
This Chinese-language README contains core descriptive content in English, and the trigger phrases are also English-only, which can impose a language expectation without explicit user opt-in. That may conflict with locale expectations for a zh-CN skill document.
