Security audit
Security checks for vulnerabilities and agentic risk
This is a Markdown-only session-management skill that may summarize or refocus conversation context, but it does not install code, access files, use credentials, call networks, or persist changes outside the chat.
Install only if you want the assistant to help summarize, reset, and narrow long conversations. For important work, ask it to preview summaries or state resets before relying on them so useful context is not accidentally dropped.
The natural-language description and the rest of the skill instructions are presented in Chinese, which effectively imposes a language requirement on users. The file does not offer an opt-in language choice or explain that the skill is intentionally limited to a Chinese-speaking or region-specific context.
The activation trigger '精简上下文' is broad enough to be invoked during normal user conversation, which can cause the skill to activate unintentionally and alter session behavior without clear user intent. In a state-management skill that prunes context, accidental activation can lead to loss of important conversational history, degraded reliability, or bypass of expected interaction flow.
The state-reset trigger conditions ('任务切换 / 迷失方向 / 错误累积') are subjective and underspecified, allowing the skill to reset state based on ambiguous signals rather than explicit authorization. In a skill designed to manage session state, this can cause premature resets, loss of task continuity, and confusion in multi-step workflows where state integrity is critical.
The attention-focus trigger conditions are vague, especially around when a task is 'too complex' or when multiple issues are being handled, which makes activation boundaries unclear. This ambiguity can let the skill suppress relevant context or redirect the conversation in ways the user did not intend, reducing transparency and increasing the risk of important information being ignored.
No suspicious patterns detected.
