The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script (scripts/extract_reassemble.py) to handle file parsing, translation extraction, and reassembly. The execution is bounded to the workspace and a temporary work directory.
[PROMPT_INJECTION]: The skill processes workspace content (Markdown and JSON) for localization. It mitigates indirect prompt injection risks by extracting prose into a structured JSON payload and using explicit instructions to the model to only translate values, preserving the structure and invariants wrapped in sentinel characters.
[EXTERNAL_DOWNLOADS]: The skill references the author's official GitHub repository for assets and installation instructions, which is standard for skill distribution. No unauthorized or suspicious external connections were identified.

l10n-sync