Security audit
Security checks for vulnerabilities and agentic risk
This AI news skill is mostly coherent, but it automatically sends generated reports to Feishu without a clear per-run confirmation step.
Review the Feishu delivery behavior before installing. Only use this skill if you are comfortable with generated reports being saved locally and sent to the configured Feishu recipient; verify the open_id and consider requiring manual confirmation before each send.
The front matter says the skill should trigger when the user asks about recent AI developments, but later instructions require immediate push delivery without waiting for the user. This contradiction is dangerous because it masks proactive behavior inside a seemingly user-invoked skill, increasing the risk of unconsented actions and covert outbound messaging.
The skill claims to operate when a user asks for AI news, but its workflow also instructs the agent to push the completed report to Feishu immediately without waiting for a request. This creates unauthorized outbound communication and turns a reactive summarization skill into a proactive notification mechanism, which can cause privacy, consent, and policy violations.
The description explicitly requires '输出中文摘要列表', which imposes a fixed language choice. There is no opt-in, user-selectable language setting, or documented justification that this skill must operate only in Chinese for compliance or region-specific reasons.
The skill instructs automatic local storage and immediate Feishu transmission without warning the user that content will be persisted and sent externally. Lack of disclosure and consent is risky because users may not expect data retention or third-party delivery, especially when the skill appears to be a simple news lookup tool.
Adding Feishu messaging introduces an external exfiltration/output channel that is not strictly necessary to collect and summarize AI news. Even if the content is usually public, unnecessary messaging capability expands the attack surface and can be repurposed to send unexpected content to third parties.
The skill mandates sending the report to a Feishu open_id immediately, but provides no recipient verification, no user confirmation, and no warning about external transmission. That can lead to misdelivery, abuse of a messaging channel, or silent distribution of content to unintended recipients.
This markdown file is entirely written in Chinese and frames sources and usage guidance only for Chinese-speaking users, while also mixing in English-source references without any opt-in or alternative locale guidance. Under the policy, language or locale restrictions should either be optional for the user or clearly justified as region-specific.
No suspicious patterns detected.
