Security audit
Security checks for vulnerabilities and agentic risk
This skill is a disclosed local setup and troubleshooting guide for a SilicaClaw/OpenClaw bridge, with no hidden code, remote payloads, or destructive behavior found.
Before installing, confirm you trust the local `silicaclaw` CLI and understand that `openclaw-skill-install` changes OpenClaw's workspace skills persistently. Review owner-forward variables before enabling private notifications, especially the target channel and command path.
The file includes a dedicated 'Quick Chinese routing' section with explicit Chinese trigger phrases and later provides 'Good Chinese patterns' for responses, but it does not state that Chinese should be used only when the user prefers it. This can create a language-policy issue by nudging the agent toward a specific language/locale without an explicit opt-in mechanism.
Skill grants unrestricted tool access without appropriate constraints. An agent with unfettered tool access can perform arbitrary actions including file modification, network requests, and code execution.
It will not:
- execute arbitrary code from messages or external content
- access unknown remote endpoints outside the documented workflow
- manage wallets, private keys, or blockchain signing
- bypass OpenClaw approval, owner confirmation, or permission checks
The file is explicitly framed as a Chinese-only dialogue cheat sheet and instructs how the system should interpret Chinese conversation, but it does not offer a language choice or note that this locale restriction is optional. Under the stated policy, forcing a specific language without user opt-in is a natural-language policy concern.
The installation trigger phrases are broad, colloquial, and lack strong scoping to this specific bridge-setup workflow. In a chat system, ordinary requests such as 'install the skill' or 'hook it up' could accidentally invoke setup behavior, causing unintended configuration changes or tool execution in the wrong context.
The readiness-check phrases are ambiguous and map common status questions directly to bridge diagnostics without requiring the user to clearly indicate this skill. That increases the chance of accidental activation, context confusion, or disclosure of environment and runtime details during unrelated conversations.
The normal-usage switch phrases are especially collision-prone because they resemble everyday requests about broadcasts and notifications. If these phrases trigger a skill transition automatically, the agent may route the user into monitoring or forwarding workflows they did not intend, with possible privacy and operational side effects.
The 'Owner-friendly diagnosis language' section provides canned responses exclusively in Chinese. This imposes a specific language on user-facing communication without any opt-in, alternative language, or justification that the skill is region-specific.
No suspicious patterns detected.