Security audit
Security checks for vulnerabilities and agentic risk
This skill is a coherent proxy tool, but it should be reviewed because it can route user web requests through third-party proxy nodes without strong warnings or tight activation boundaries.
Install this only if you intentionally want an agent to create and use third-party SOCKS5 proxies. Avoid sending passwords, session cookies, personal data, private network URLs, or sensitive API traffic through the proxy, and treat generated proxy credentials as secrets. Configure agents to ask before proxying requests and to keep use within applicable laws and site terms.
The description lists generic keywords such as '代理', '工具', 'api', and '网络请求' as applicable triggers. These terms are broad and common in normal developer conversations, and the file does not provide clear negative examples or explicit constraints on when the skill should or should not activate.
The skill routes user-specified HTTP traffic through third-party proxy infrastructure but does not prominently warn that destinations, headers, and potentially sensitive content may transit or be observable by external network operators. In this context, the risk is elevated because the skill is explicitly designed for proxy-based request forwarding and geolocation evasion, so users may unknowingly send sensitive requests through untrusted intermediaries.
The '能力覆盖范围' statement includes fragmented, non-specific phrases like '通过去中心化' and '代理发起网络请求', which do not clearly define invocation boundaries. Without concrete trigger phrases or exclusion conditions, the skill may be selected for unrelated requests involving networking, regions, or proxies.
The skill warns against echoing or logging secrets, yet its example output includes full proxy credentials. In practice, users and agent implementations often copy example output patterns into logs, chat responses, or telemetry, which can leak live or reusable proxy access and enable unauthorized use of paid network resources or attribution abuse.
The description claims '去除外部依赖引用', implying reduced or removed external dependencies. However, the dependency section later states that curl is required, jq is recommended, and the proxy network API is mandatory. This is a direct mismatch between the stated differentiator and the documented implementation requirements.
The capability table explicitly marks '请求重试与熔断' as unsupported in the free version. Later, the skill documents recovery behavior such as retrying after service unavailability, which presents retry behavior as part of the skill's operation. This creates a user-facing mismatch about what the free edition actually does.
