The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes subprocess.run to execute system commands necessary for macOS orchestration and file management:
Executes osascript to run AppleScript and VBA macros, allowing the agent to control Microsoft Excel for tasks like refreshing pivot tables and performing calculations.
Uses the open command to launch generated Excel files for the user.
Employs xattr -d com.apple.quarantine to remove the macOS quarantine attribute from generated workbooks, ensuring they can be opened without repetitive security warnings.
[EXTERNAL_DOWNLOADS]: The skill instructions include steps to install well-known and trusted Python libraries required for its primary functionality:
openpyxl, Pillow, and pandas are installed via pip. These are standard industry tools for Excel manipulation and data processing.
[DYNAMIC_EXECUTION]: The skill generates and executes code at runtime to interface with the Excel application engine:
Dynamic AppleScript and VBA strings are constructed based on the workbook structure and executed via the shell. This is a standard pattern for macOS application automation where fixed libraries are insufficient.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from Excel workbooks, which represents a potential ingestion surface for indirect prompt injection:
Ingestion points: Loading and reading existing .xlsx and .xlsm files using openpyxl in SKILL.md and openpyxl-reference.md.
Boundary markers: There are no explicit instructions to the agent to treat cell contents as untrusted data or use specific delimiters when interpolating data into its logic.
Capability inventory: The skill provides the agent with capabilities to execute shell commands and modify file system attributes.
Sanitization: The skill includes an extensive quality audit system (audit-system.md) that checks for data integrity and formatting errors, but it does not specifically sanitize input for malicious prompt patterns.

xlsx-design-agent