The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill's installation process and runtime bootstrap wrapper download the excelcli binary runtime from GitHub releases (github.com/sbroenne/mcp-server-excel).
[COMMAND_EXECUTION]: The skill is designed to execute local shell commands using the excelcli tool to automate Excel workbook operations.
[DYNAMIC_EXECUTION]: The skill facilitates the execution of dynamic code within the Excel environment, including VBA macros (vba run), Power Query M code (powerquery evaluate), and Microsoft 365 Python in Excel formulas (pythoninexcel set-formula).
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from Excel workbooks (e.g., via range get-values or table get-data). It lacks explicit boundary markers to prevent the agent from following instructions embedded within the workbook data. The agent's extensive capabilities (shell execution, VBA, file writing) increase the potential impact of such an injection.
[PROMPT_INJECTION]: The skill includes behavioral directives that emphasize agent autonomy, such as "NEVER Ask Clarifying Questions" and "Execute tasks immediately without asking for confirmation," which could potentially override base agent safety guidelines regarding user oversight.
[DATA_EXFILTRATION]: The skill documentation mentions the ability to send DAX and Power Query M code to external formatting services (daxformatter.com and powerqueryformatter.com), though it instructs the agent to obtain user consent before doing so due to privacy considerations.

excel-cli