Security audit
Security checks for vulnerabilities and agentic risk
This is a coherent coding-agent skill, though users should be aware it can route broad coding requests to sub-agents with file-reading and command-execution tools.
Install this if you want a Chinese-language coding workflow with review, security, and testing sub-agents. Use it in repositories where you are comfortable allowing file inspection and project commands, and watch for accidental activation from broad terms like review, security, or test.
The template specifies display_name: "显示名称" and describes it as a Chinese-friendly name, while the overall document is written as the standard agent definition. This imposes a specific language/locale expectation in the skill definition without indicating user opt-in or a documented regional constraint.
The routing logic activates the coding agent for broadly defined programming-related requests and falls back to a default reviewer when no clear match exists. In agent systems, overly broad activation increases the chance that this skill is invoked outside its intended scope, which can expose code context, trigger unnecessary sub-agents, or cause unintended tool-enabled processing of user input.
Trigger examples like 'review', 'security', and 'test' are common everyday terms and are not scoped to code-related contexts. This can cause accidental or adversarial invocation of specialized agents, leading to misrouting, unnecessary access to repository context, and downstream execution paths if child agents are configured with powerful tools.
Natural-language instructions, labels, and output requirements in the file are all specified in Chinese, and the manifest does not state that the skill is region-specific or provide an opt-in language selection. This can amount to a language policy violation if users are implicitly forced into a specific language without consent.
The trigger list includes broad, everyday phrases such as '帮我看看这个文件' and '这段代码怎么样' that can match loosely phrased user requests and invoke the skill unintentionally. Because this agent has access to Read, Grep, Glob, and exec, accidental invocation can cause unnecessary file inspection or command execution in contexts where the user did not explicitly request code review.
The trigger phrases are broad enough that the skill may activate during ordinary discussion about security rather than an explicit request to run a specialized audit workflow. Because this skill has access to powerful tools including exec, over-broad activation increases the chance of unintended tool-enabled analysis or command execution in contexts where the user only wanted conversational guidance.
The skill is written entirely in Chinese and prescribes Chinese-language behavior without any visible locale negotiation or user opt-in. This is primarily a safety and usability issue: it can cause misunderstanding of instructions, test output, or remediation guidance for users expecting another language, which may lead to incorrect actions or missed warnings.
The trigger examples include very broad phrases such as '写测试', 'unit test', and '补充测试', which are common requests and can cause the skill to activate in situations where the user did not specifically intend to invoke this agent. In a tool-enabled skill with exec permission, unintended activation increases the chance of unnecessary code inspection or command execution in the wrong workflow context.
The field description for display_name requires a Chinese-friendly name, and the document's normative templates are presented only in Chinese. This can amount to a language policy constraint without offering users or implementers an explicit language or locale choice.
The spec defines a catch-all fallback to code-reviewer whenever no trigger matches, which broadens activation beyond explicit user intent. In an agent-routing system, this can cause unintended handling of arbitrary inputs by a powerful default agent, increasing the chance of prompt injection exposure, overreach, or misuse of tools if that default agent has broad capabilities.
The manifest's natural-language content, display name, description, trigger examples, and system prompt are all written solely in Chinese, with no indication that users may choose another language. This can violate language/locale policy when a skill implicitly constrains interaction language without opt-in or documented justification.
No suspicious patterns detected.
