Security audit
Security checks for vulnerabilities and agentic risk
This is a documentation-only workflow helper with broad activation wording, but I found no hidden execution, data access, persistence, or malicious behavior.
This appears safe to install as a lightweight planning helper, but expect possible unintended activation on generic PDF or editing requests. Use explicit invocation for this skill, and disable or narrow implicit invocation if your platform allows it.
The manifest description says to use the skill when a user asks for broad terms like "pdf," "edit," or a "practical workflow," which are common across many benign tasks. This ambiguous activation guidance can make the skill match far outside its intended scope, leading to misrouting and unintended exposure of the skill's instructions in unrelated sessions.
The trigger list includes highly generic terms such as "nano," "pdf," "edit," "natural," and "language," which can cause the skill to activate in many unrelated conversations. Over-broad activation increases the chance of inappropriate routing, prompt-context pollution, and accidental application of the skill in contexts it was not designed to handle.
触发关键词包含大量高频、通用词，如“nano”“pdf”“edit”“natural”“language”“instructions”“bug fix”，且示例触发句也非常宽泛，容易在无关上下文中误触发该技能。这会导致技能被过度调用、抢占本应由更精确技能处理的任务，并可能把与 PDF 无关的请求错误引导到该工作流。
The default prompt trigger is broad and composed of common terms like "help me," "work productivity," and "pdf," which are likely to appear in ordinary user requests. Combined with implicit invocation, this can cause the skill to activate unintentionally, giving the skill undue influence over conversations and increasing the chance of prompt-routing or context-injection issues.
The trigger sentence begins with a very broad everyday phrase ('Help me ...'), which can cause the skill to activate during ordinary conversation even when the user did not intend to invoke this specific skill. In an agent environment, over-broad activation increases prompt-routing mistakes, causing the wrong workflow to run on unrelated requests and potentially exposing downstream tools or actions unnecessarily.
The trigger phrases are broad enough to match ordinary requests involving PDFs, editing, workflows, or generic implementation help, which can cause the skill to activate outside its intended scope. In an agent ecosystem, overbroad invocation increases the chance of prompt-routing mistakes, unintended instruction injection into unrelated tasks, and user confusion about why this skill was selected.
The trigger phrases are broad enough to match ordinary requests involving PDFs, editing, workflows, or general implementation help, which can cause the skill to activate outside its intended scope. In an agent ecosystem, this creates prompt-routing risk: users may be steered into this skill unexpectedly, leading to incorrect handling, policy bypass opportunities, or confusion about what instructions are governing the session.
The example trigger sentences are vague, repetitive, and do not clearly distinguish valid invocations from ordinary user requests. Poor trigger examples reinforce over-broad matching behavior and make it harder for maintainers or routing systems to enforce safe, predictable activation boundaries.
技能描述中的调用条件表述为“当用户提出 work-productivity, nano-pdf, nano, pdf, edit，或需要围绕该需求获得实用流程……时使用”，边界较模糊，未清晰说明何时不该调用。模糊的适用范围会放大误匹配概率，尤其在多技能环境中，可能造成任务路由混乱、输出偏题或错误依赖该技能。
This trigger sentence is ambiguous about when the skill should activate and does not clearly bound its scope, making accidental invocation more likely. Because the skill is framed broadly around practical workflow help, ambiguous routing can lead to irrelevant or unintended execution paths, reducing reliability and increasing the chance of misuse in multi-skill systems.
In a Chinese-language README, the core usage guidance and trigger phrases are presented in English, which implicitly steers interaction toward a specific language without stating that users may choose their preferred language. This can be a language/locale policy issue because no opt-in or justification for the English-only usage pattern is provided.
No suspicious patterns detected.
