Security audit
Security checks for vulnerabilities and agentic risk
This is a documentation-only orchestration template that guides sub-agent delegation and review, with some routing and persistence cautions but no evidence of hidden or malicious behavior.
Install this if you want a main agent to coordinate OpenClaw sub-agents. Review the routing keywords, language of templates, allowed sub-agent list, and any MEMORY.md or resume-state use before enabling it in a shared or mixed-language workspace.
The trigger list includes broad phrases such as generic orchestration terms that can match ordinary user requests, causing the skill to auto-activate outside its intended scope. In an orchestrator skill, unintended activation is more dangerous because it can reshape task routing, delegation, and acceptance behavior for unrelated conversations.
The embedded reporting and delivery templates prescribe Chinese-language output fields and labels without indicating that locale is configurable or user-driven. In an orchestrator skill, hard-coded language can cause operator confusion, misrouting, failed acceptance checks, or accidental disclosure through misunderstood status/escalation messages, especially in mixed-language environments.
The routing examples are entirely written as Chinese user utterances, which implies a language-specific interaction pattern for activation and use. The document does not state that this playbook is intentionally Chinese-only or provide any user opt-in or alternative language handling, creating a natural-language locale policy concern.
The file defines routing triggers and examples largely in Chinese, mixed with a few English phrases, but does not state that the skill is intended only for Chinese-language workflows or provide an opt-in language selection. This can violate language/locale policy because it effectively biases activation behavior toward one language without user choice.
The no-spawn trigger phrase "What is" is overly broad and can match many requests that are not simple factual questions. This can force local handling when delegation or bounded specialist review would be safer or more appropriate, increasing the chance of incorrect routing and missed verification in an orchestration system.
Broad domain triggers like "内容", "功能", and "投资" are likely to overmatch everyday language and route tasks into inappropriate domains. In an agent orchestration context, misclassification can send work to the wrong specialist, weaken safety assumptions around bounded delegation, and increase the risk of unintended side effects or poor handling of sensitive tasks.
Line L123 lists Chinese phrases as part of the no-spawn rules, but the document does not state that the skill is Chinese-language or otherwise locale-specific. This creates a natural-language locale bias because special handling is defined for one language without user opt-in or an explicit multilingual policy.
The manifest presents category, title, descriptions, and triggers in Chinese in some sections while other sections are in English, but it does not explain locale selection, user opt-in, or whether the skill is intended for a specific language audience. This can create an implicit language policy constraint without clear documentation or user choice.
No suspicious patterns detected.
