The Agent Skills Directory

[DATA_EXPOSURE]: The skill accesses absolute file paths of projects within the user's workspace.
Evidence: The get_workspace_info tool is called to retrieve project metadata, which includes absolute paths (e.g., /absolute/path).
Context: This data is used internally by the skill to correctly scope configuration listing and execution via the manage_run_configuration tool.
[COMMAND_EXECUTION]: The skill is designed to trigger the execution of Mule applications.
Evidence: It uses the manage_run_configuration tool with the execute operation to start processes in run or debug mode.
Context: This functionality is the primary intended use case for the skill provided by the developer tooling.
[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection through the processing of untrusted project or configuration names.
Ingestion points: User-provided project names, configuration names, and workspace metadata retrieved from the environment.
Boundary markers: None present; the agent is instructed to use names directly from the workspace or user input.
Capability inventory: Listing projects, creating configurations, and executing run/debug processes.
Sanitization: No specific sanitization or validation of project/configuration names is mentioned before they are passed to tool parameters.

execute-mule-run-config