The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions and a manual testing script that uses subprocess.Popen to execute server commands and pkill to terminate orphaned processes.
[EXTERNAL_DOWNLOADS]: The skill utilizes npx -y and uvx to fetch and run MCP server packages. It specifically references official packages within the @modelcontextprotocol scope, which is recognized as a well-known service.
[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection because it instructs the agent to analyze application logs and configuration files that may contain data from untrusted external sources.
Ingestion points: ~/Library/Logs/Claude/mcp*.log and claude_desktop_config.json.
Boundary markers: No explicit delimiters or boundary markers are provided for the agent when reading log files or configuration data.
Capability inventory: The skill has the capability to execute system commands via subprocess, manage processes with pkill, and read/write file system data.
Sanitization: There is no evidence of sanitization or validation performed on the contents of logs or external configuration files before they are processed by the agent.

mcp-server-orchestrator