Security audit
Security checks for vulnerabilities and agentic risk
This is a documentation-only workflow helper with overly broad activation wording, but no hidden code, persistence, credential use, or destructive behavior.
Install only if you want a general helper for Skill Vetter-style workflow planning. Be aware it may trigger on unrelated security, GitHub, or bug-fix requests because its activation terms are broad.
The manifest description includes very broad activation terms such as "security" and "first," which can match many ordinary user requests unrelated to this skill’s narrow purpose. In an agent-routing context, this can cause inappropriate auto-activation, expanding the skill’s effective authority and creating prompt-scope confusion or unsafe workflow injection into unrelated tasks.
The trigger keyword list contains ambiguous everyday terms like "security," "first," "before," and "github," which are likely to appear in many benign conversations. This makes accidental invocation much more likely, potentially hijacking unrelated workflows and causing the agent to apply this skill in contexts where its instructions are irrelevant or disruptive.
The trigger keywords are overly broad and include generic terms like "security", "first", "before", "github", and "bug fix", which can cause this skill to activate for many unrelated requests. In an agent environment, this creates routing confusion and increases the chance that users receive the wrong workflow or that a broad skill intercepts more specialized safety-sensitive tasks.
The trigger sentences and keywords are broad enough to match many ordinary requests such as generic security, bug-fix, GitHub, or workflow help, which can cause the skill to activate outside its intended scope. In an agent ecosystem, overbroad activation can silently route users into the wrong workflow, creating confused-deputy behavior, poor security vetting outcomes, and accidental handling of unrelated tasks under misleading assumptions.
The trigger examples are broad enough to match common requests like general security help, GitHub help, or bug fixing, which can cause this skill to activate outside its intended scope. In an agent ecosystem, over-broad routing can divert sensitive or unrelated tasks into a generic workflow, producing incorrect handling, misleading outputs, or interference with more appropriate specialized skills.
The trigger phrases are broad enough to match ordinary requests about productivity, security, GitHub, or bug fixing, which can cause the skill to activate outside its intended scope. In an agent ecosystem, overbroad activation can route unrelated user tasks into this workflow, leading to confused delegation, unexpected processing of sensitive context, or bypass of more appropriate safety/review paths.
The example trigger sentences reinforce the same vague activation pattern by demonstrating invocation from broad, loosely phrased requests without any boundary conditions. This trains routing or downstream authors toward overbroad use, increasing the chance of accidental activation, though the impact is somewhat lower than the manifest and keyword issues because examples are secondary signals.
The file is a zh-CN skill variant and does not present a user language selection or fallback mechanism within the skill content. This can cause user misunderstanding, mis-execution of instructions, or poor safety outcomes if the agent applies the skill to users who do not understand Chinese.
The description says to use the skill when users ask for broad categories like work-productivity, security, or practical workflow help, but it does not clearly define scope boundaries. This ambiguity can lead to accidental invocation on unrelated requests, especially because "security" and artifact/checklist requests are common across many skills.
The default prompt embeds a very broad invocation phrase tied to common terms like work-productivity, security, analysis, checklist, and implementation support, which can overlap with ordinary user requests. Combined with allow_implicit_invocation: true, this increases the chance the skill is invoked unintentionally, causing prompt injection surface expansion, unexpected behavior, or disclosure of skill behavior in contexts where the user did not explicitly request it.
This file is presented as a Chinese README, but substantial user-facing sections and trigger text remain in English, including the core description and invocation examples. That can amount to a language-policy issue because it implicitly forces English content on users of the Chinese-localized document without any opt-in or stated language preference.
