The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the subprocess module to run internal checks and external tools like pdftotext and pandoc. These calls use argument lists and are protected by a safety utility that prevents directory traversal.
[INDIRECT_PROMPT_INJECTION]: The skill processes manuscript data for auditing purposes. Ingestion points include manuscript and abstract files, where content is used for word counting and pattern matching. Capability inventory includes file-write and subprocess calls, but these are driven by script logic rather than user input.
[DYNAMIC_EXECUTION]: On macOS, the skill drives Microsoft Word using dynamically generated AppleScript. Author names and file paths are escaped before interpolation into the script to mitigate injection risks.

sync-submission