The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute project-specific development tools including breeze and a tool named prek (likely a typo for the standard pre-commit framework). These commands are used for checking translation completeness and running validation hooks.
breeze ui check-translation-completeness --language <locale>
prek run --from-ref main --hook-stage pre-commit
[INDIRECT_PROMPT_INJECTION]: The skill processes external data in the form of translation strings stored in JSON files, creating an indirect prompt injection surface.
Ingestion points: Translation JSON files located in airflow-core/src/airflow/ui/public/i18n/locales/.
Boundary markers: The instructions explicitly warn not to translate or remove variable names inside {{…}} delimiters, which helps preserve the integrity of the interpolation logic.
Capability inventory: The skill uses breeze and pre-commit hooks to validate and format files, but does not perform dangerous operations like arbitrary code execution or network exfiltration based on the ingested content.
Sanitization: The skill relies on pre-commit hooks and the breeze tool's completeness check to validate the structure and content of the translation files.

airflow-translations