Security audit
Security checks for vulnerabilities and agentic risk
This is a documentation-only unit-test coverage helper with broad activation wording but no hidden execution, persistence, credential use, or unrelated data access.
Install this if you want repeatable help improving unit tests and coverage. Be aware it may be invoked by broad testing-related phrases, so prefer explicit use when you want this workflow and review any proposed code changes as you normally would.
The trigger sentence is so generic that it can activate on ordinary user phrasing rather than a clear, intentional invocation of this skill. That can cause the wrong skill to be selected, steering unrelated requests into test-coverage workflows and creating prompt-routing confusion that an attacker could exploit to hijack task handling or suppress more appropriate safeguards.
The trigger phrases are generic and can match broad user requests about testing or practical workflows, which increases the chance the skill activates outside its intended scope. Over-broad activation can cause incorrect routing, unexpected disclosure of project context to the wrong skill, or unsafe reliance on irrelevant guidance, even though it does not directly execute code.
The trigger phrases are broad and template-like, including generic terms such as software-and-data, testing, regression, and practical workflow. This increases the likelihood the skill is invoked in unintended contexts, which can lead to scope confusion, accidental activation, and unsafe delegation of tasks to a skill that may operate on codebases without the user explicitly requesting it.
The skill description is broad enough to match many ordinary software testing or quality-related requests, which can cause the agent to invoke this skill when a more specific or safer skill would be more appropriate. Over-broad routing is dangerous because it increases unintended activation, expands the skill’s effective authority, and can interfere with correct task selection in multi-skill environments.
The trigger keywords include very generic terms like 'testing', 'regression', and 'quality', which are common across many unrelated developer requests. In an automated skill-selection system, such generic triggers can cause accidental invocation and prompt hijacking of normal requests, reducing reliability and potentially exposing the user to the wrong workflow or outputs.
L04 说明“当用户提出 software-and-data, unit tests, test coverage, testing, regression，或需要围绕该需求获得实用流程、产物、检查清单、分析或实现支持时使用”，其中包含如“testing”这类宽泛主题词，并用“或需要…支持”作兜底条件，未清楚限定何时应激活、何时不应激活。这种描述容易与大量日常开发/测试相关请求重叠，导致非预期调用。
L44 的关键词包含“testing”“quality”等高频通用词，L46-L50 只给出了正向示例，没有说明这些词在何种上下文下不应触发技能。对于 markdown 技能说明，这会造成触发范围过宽，与普通质量讨论或测试对话发生碰撞。
The default prompt embeds a very broad natural-language trigger phrase ('help me' plus generic testing terms) while implicit invocation is allowed, which can cause the skill to activate in many ordinary software-assistance conversations. This increases the risk of unintended routing, prompt-scope confusion, and accidental invocation in contexts the user did not explicitly choose.
The file lists separate English and Chinese instruction and guide files, but does not state how the user's language preference is selected or whether the user can opt in to a locale. This can indicate a locale-handling policy gap if the skill defaults to a language without explicit user choice.
The example trigger beginning with 'Help me' is too unconstrained and reinforces a permissive invocation pattern for a broad class of requests. While less severe than the keyword issue, it still contributes to ambiguous routing behavior and increases the chance the skill is selected without sufficient evidence that the user actually wants unit-test coverage assistance.
