The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/numbers_tools.py dynamically constructs AppleScript code by interpolating user-controlled variables (such as file paths, sheet names, and table data) into template strings. This code is then executed using subprocess.run(['osascript', '-e', ...]). This pattern of building executable scripts from external input is a high-risk practice that can lead to command injection within the AppleScript environment.
[PROMPT_INJECTION]: The skill features an attack surface for indirect prompt injection as it processes complex JSON data from external sources that directly influences the automation logic.
Ingestion points: CLI arguments in scripts/numbers_tools.py including headers_json, rows_json, workbook_json, sheet_json, and style_json.
Boundary markers: None are employed within the script's interpolation logic to delimit untrusted data.
Capability inventory: The skill can interact with the file system and execute system commands (osascript and mdls) via subprocess.run as implemented in scripts/numbers_tools.py.
Sanitization: The script uses a basic _esc function in scripts/numbers_tools.py that only escapes backslashes, double quotes, and newlines. This minimal sanitization may not be sufficient to prevent all forms of AppleScript injection.

apple-flow-numbers