The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run and subprocess.Popen within its Python control plane to manage internal components, including a C# supervisor binary, the .NET CLI, and Git. These invocations are implemented using list-based arguments for safety and are scoped to the skill's local environment.
[REMOTE_CODE_EXECUTION]: The skill involves building and executing local C# binaries to drive Excel COM automation. This execution is monitored by a supervisor process that uses Windows Job Objects to ensure rigorous process cleanup and prevent orphaned instances.
[PROMPT_INJECTION]: The skill ingests external data from workbooks and CSV files for validation and processing. Indirect prompt injection surfaces are mitigated by processing data through established libraries (openpyxl, xlsxwriter) and returning results as structured JSON, ensuring that data values are not interpreted as instructions.
[CREDENTIALS_UNSAFE]: The skill implements a security redaction layer in control_plane/audit_manifest.py that identifies and removes secrets such as connection strings, API keys, and bearer tokens from results before they are written to the audit log.

xlsx-win