Security audit
Security checks for vulnerabilities and agentic risk
This is a transparent context-usage diagnostic skill, with only a minor caveat that some trigger words are broad enough to activate unintentionally.
Use explicit phrases like “context doctor” if you want to avoid accidental activation. Before installing, be comfortable with the agent summarizing session metadata such as token usage, model/cost data, tool-call counts, and large-output rankings; avoid enabling heartbeat auto-checks unless you want periodic reports.
The trigger phrase "诊断" is very broad in normal conversation and can easily appear in unrelated user requests, causing unintended activation of the skill. In a skill that inspects session state and reports token/context usage, accidental invocation can waste context, reveal conversation metadata, and interfere with the agent's intended task flow.
The phrase "checkup" is generic English vocabulary and may be used in ordinary conversation unrelated to context diagnostics. This raises the risk of unintentional skill execution, which can consume tokens, interrupt workflow, and expose session-level metadata when the user did not intend to request it.
The trigger list includes very generic phrases like “诊断” and “健康检查”, which can match ordinary conversation and cause the skill to activate outside the user’s intended scope. In an agent setting, ambiguous activation can lead to unnecessary introspection about session state, extra token consumption, and accidental disclosure of operational metadata when the user did not explicitly request this skill.
The file consistently specifies the skill in Chinese and presents the report format, labels, and operating guidance only in Chinese, with no indication that the response language should follow user preference. This can create a locale-policy issue if the organization expects language choice or user opt-in rather than a fixed language.
The usage guidance repeats broad activation conditions such as “体检” and “诊断” without defining boundaries, reinforcing accidental triggering during normal conversation. Because this skill inspects session/tool usage state, mistaken activation can consume more context and expose conversation-structure metadata that may not be relevant or desired.
The documentation is presented only in Chinese and English, with no indication that the skill supports or adapts to the user's preferred language or locale. For a language/locale policy review, this can be read as an implicit two-language constraint without opt-in or justification.
No suspicious patterns detected.
