The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's functionality is centered on the execution of the excelcli command-line utility. It constructs and runs shell commands with various parameters to perform Excel operations.
[EXTERNAL_DOWNLOADS]: Installation instructions include the use of dotnet tool install --global Sbroenne.ExcelMcp.CLI and npx skills add sbroenne/mcp-server-excel. These are vendor-provided resources from 'sbroenne'.
[REMOTE_CODE_EXECUTION]: The vba command group allows the agent to import, update, and run VBA procedures (excelcli vba run --procedure-name ...). This enables the execution of arbitrary code within the Excel process on the host machine.
[PROMPT_INJECTION]: Instructions in SKILL.md and behavioral-rules.md direct the agent to bypass confirmation steps and avoid asking clarifying questions (e.g., 'NEVER Ask Clarifying Questions', 'Execute immediately'). This emphasis on autonomy increases the risk of unintended command execution.
[PROMPT_INJECTION]: The skill exposes an indirect prompt injection surface. Untrusted data enters the agent context through ingestion points like range(get-values) and table(get-data) (SKILL.md). There are no boundary markers or explicit sanitization processes mentioned. The agent's capability inventory includes subprocess execution, VBA macro running, and file-write operations, creating a risk if the agent obeys instructions embedded in processed Excel files.

excel-cli