The Agent Skills Directory

[SAFE]: The skill defines a set of rules for Chinese translation and localization across three regional variants (CN, TW, HK). It focuses on script consistency, vocabulary differences, and regional formatting standards. The instructions are designed to ensure native-quality output and prevent common localization errors.
[SAFE]: The instructions regarding "Political Red Lines" are designed as safety guardrails for the agent to avoid generating content that could lead to legal or brand-reputation issues in specific jurisdictions. They include instructions to seek user confirmation before proceeding with sensitive content, which serves as a human-in-the-loop safety mechanism.
[PROMPT_INJECTION]: The skill processes external source text for translation, which represents an indirect prompt injection attack surface.
Ingestion points: Source text for translation is passed into the agent context from the user.
Boundary markers: The skill lacks explicit technical delimiters but provides specific keywords for the agent to monitor and instructions to halt execution for manual confirmation if sensitive terms are detected.
Capability inventory: The skill instructs the agent to use standard UI tools (AskUserQuestion) but does not perform file writes, network requests, or subprocess execution.
Sanitization: The skill mandates character script validation (Simplified vs Traditional) and explicit user confirmation for sensitive terms, acting as a validation and sanitization step.

localize-zh