The Agent Skills Directory

[SAFE]: The skill is a utility designed to compress Japanese output for token efficiency. It provides detailed rules for removing honorifics, particles, and redundant words while explicitly instructing the AI to maintain technical accuracy.
[COMMAND_EXECUTION]: Static analysis flags the presence of destructive command strings such as rm -rf / and DROP TABLE. However, these are contextually safe as they are defined as triggers for 'Auto-Clarity'—a security feature that forces the AI to switch from compressed speech back to standard Japanese when describing dangerous operations to prevent user misunderstanding.
[PROMPT_INJECTION]: The skill contains instructions to override the default communication style. These instructions are confined to stylistic preferences and do not attempt to bypass core safety filters, ignore ethical constraints, or extract internal system prompts.
[DATA_EXFILTRATION]: No network activity, sensitive file path access, or credential harvesting patterns were identified within the skill instructions.
[EXTERNAL_DOWNLOADS]: The skill does not reference or attempt to download any external packages, scripts, or remote resources.

genshijin