Security audit
Security checks for vulnerabilities and agentic risk
This skill provides OpenAPI documentation guidance and I found no hidden execution, persistence, or unsafe data handling.
Install this for OpenAPI or Swagger documentation tasks. Because the trigger metadata is somewhat broad, users should make sure it is only invoked for API documentation work and should avoid including real secrets, customer data, or live credentials in examples.
The skill description is broad enough to match common requests about API docs, schemas, reviews, and workflows, which increases the chance of accidental or overly frequent activation. Overbroad triggering can cause the wrong skill to handle user requests, leading to unnecessary access to repository context, misleading outputs, or reduced reliability of agent behavior.
The keyword list includes highly generic terms such as 'schema' and 'SDK', which can appear in many unrelated contexts and may spuriously activate the skill. This creates prompt-routing risk: unrelated requests may be handled as OpenAPI work, causing confusion, poor task execution, or unintended exposure of contextual project information to an unnecessary skill path.
文件名和全文内容表明该技能说明固定为简体中文版本，但文档中没有说明这是可选语言之一，也没有提供用户语言偏好选择。根据规则，若技能强制特定语言而无用户选择或明确的地区合规理由，可构成语言/locale 策略问题。
该描述写明“当用户需要 OpenAPI、Swagger、REST API、API docs、schema 相关的流程、代码、清单、文档或审查支持时使用”，覆盖面非常广，几乎包含任何与 API 文档或 schema 相关的请求，但没有说明边界、排除条件或不应触发的场景。这种宽泛表述容易与日常技术支持对话重叠，增加非预期调用该技能的风险。
“OpenAPI、Swagger、REST API、API docs、schema、SDK、contract testing”中多项术语尤其是“schema”“SDK”属于高频泛化开发词汇，单独出现时并不足以明确应调用本技能。该部分也未提供组合条件、上下文限制或负例说明，因此触发范围不够明确。
The skill metadata uses broad trigger language and a permissive default prompt that can cause the skill to activate for a wide range of generic API-documentation-related requests. Because implicit invocation is enabled, the platform may route user tasks to this skill when the user did not explicitly request it, increasing the chance of unintended behavior or misuse of any downstream capabilities the skill has.
The trigger sentence is extremely broad and malformed ('Help me ...') such that ordinary user language could unintentionally invoke this skill outside a clearly bounded OpenAPI-documentation context. Overbroad triggers increase the chance of accidental routing, causing the agent to apply the wrong skill, mishandle user intent, or expose internal workflow behavior in unrelated conversations.
The invocation description lacks clear trigger boundaries and repeats a long generic requirement statement instead of concrete activation criteria. This ambiguity can make the orchestration layer select the skill for loosely related requests, reducing intent isolation and creating prompt-routing errors that may misdirect sensitive code, schemas, or API details into an inappropriate workflow.
该文件整体以简体中文呈现用户说明，但未在文档中说明这是可选语言版本，也未给出用户可选择其他语言的提示。按规则，若技能强制特定语言且没有用户选择或明确合理说明，属于自然语言层面的策略风险。
