The Agent Skills Directory

[PROMPT_INJECTION]: The skill converts internal task instructions into Mandarin, creating an indirect prompt injection surface where translation could potentially be used to bypass safety filters or introduce semantic ambiguity in sub-agent tasks.\n- Ingestion points: Instructions and prompts intended for sub-agents based on the active user session in SKILL.md.\n- Boundary markers: The skill does not employ explicit delimiters for translated strings, although it instructs the agent to maintain technical terms in English.\n- Capability inventory: The skill does not define access to external tools, file system modifications, or network resources.\n- Sanitization: There is no explicit logic provided to sanitize or verify the safety of the translated content before dispatch.

china-maxing