The Agent Skills Directory

[SAFE]: The skill functions as a template-based text transformer. It does not perform network operations, execute shell commands, or attempt to persist across sessions.
[DATA_EXPOSURE]: The skill references a local file path (../../prompts/data/cross-model-matrix.json) to retrieve its benchmark data. This access is scoped to the project's own data assets and does not involve harvesting sensitive system files or user credentials.
[PROMPT_INJECTION]: The skill ingests user-provided video prompts for translation. It manages this risk by using structured few-shot templates and clear delimiters, which prevents the user input from being interpreted as a command to the underlying agent. There is no evidence of instructions that attempt to bypass AI safety guidelines.

prompt-translator