The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input, specifically project selections and custom configuration names, which are then used as parameters for the manage_run_configuration tool. This tool performs file-write operations to create .code-workspace or launch.json files. While the instructions state that the tool handles validation, this input surface could potentially be used for schema confusion or path manipulation if the underlying tool's sanitization is bypassed.
Ingestion points: User-provided configName and project selection strings collected in Step 2 and Step 3 of the process.
Boundary markers: Absent. The workflow is structured with numbered lists and explicit options, but the instructions do not explicitly tell the agent to sanitize or ignore embedded control characters in the user-supplied configuration name.
Capability inventory: The skill has the capability to write configuration files and execute existing configurations through the manage_run_configuration tool.
Sanitization: The skill instructions note that the tool handles all validation and file operations, delegating sanitization to the underlying tool implementation rather than handling it within the prompt logic.

create-mule-run-config