Security audit
Security checks for vulnerabilities and agentic risk
This is a disclosed workflow-composition skill that inventories available experts and helps coordinate them, with no evidence of hidden execution, credential use, persistence, or exfiltration.
Before installing, be comfortable with the skill listing available marketplace and locally installed experts/teams and using that inventory to suggest or coordinate workflows. Review any custom files you add under references/, keep sensitive business data out of reusable templates, and require explicit confirmation before allowing any write, publish, delete, connector, or expert-dispatch action.
The skill declares its own contents must not be modified, but then explicitly allows users to add and register files under references/. This creates a policy contradiction that can be abused to smuggle untrusted prompt content into later executions, especially because the skill instructs the agent to load references/ files into context and use them as authoritative guidance.
The skill mandates bilingual user-facing output as a fixed policy rather than adapting to the user's stated language or preference. This can cause unnecessary disclosure, confusion, or prompt-surface expansion in downstream workflows, especially when concise or single-language responses are expected.
The validation rule hard-requires bilingual copy without offering a language choice, which overrides user intent and can degrade safety and usability in multilingual settings. In agent workflows, rigid output policies can also interfere with downstream tools or specialists that expect one language only.
All user-facing headings, inputs, outputs, and guidance in this template are presented exclusively in Chinese, and there is no indication that users may choose another language or locale. Under the policy, forcing a specific language without user opt-in is a natural-language policy concern.
The requirement that all user-facing copy be bilingual Chinese-English is a hard-coded output policy that overrides user preference and can cause unnecessary disclosure or confusion, especially when prompts or generated content may include sensitive business context that the user only intended to express in one language. In this skill, the risk is primarily policy and UX-related rather than direct code-execution risk, but it is still a real issue because the instruction is mandatory and repeated across the workflow.
This markdown skill reference is entirely presented in Chinese and does not provide an opt-in, alternate language, or note that the skill is intended only for Chinese-speaking users. Under the policy, forcing a specific language without user choice can be a natural-language policy violation.
Line L134 requires output language to be consistent with the user's original request. While narrower than a fixed-language mandate, it still enforces a locale policy without explicitly allowing the user to request another language or bilingual output.
The natural-language instruction includes '输出语言与用户一致' as a fixed rule. Under the policy, language constraints can be a violation when they are enforced without explicit user choice or documented justification.
No suspicious patterns detected.
