This server has severe definition quality issues. Tool 1 (execute_skill_action) has a generic, uninformative description (47 chars) that violates the 10-1024 char guideline by being too vague. Tool 2 (synthesize_flux_diffusion_frame) has a reasonable description (127 chars) and explicit input schema with types and descriptions, but the server's mcp_server.py does NOT actually register tools via the MCP protocol, it only prints a JSON status message. Tool definitions appear to exist only in client.py as a Python class, not as MCP tool registrations. Additionally, there is no output schema documentation for either tool, and error handling is absent. The server does not implement the MCP protocol correctly.

• Implement proper MCP tool registration in mcp_server.py using the MCP SDK. Each tool must be registered with a name, description, and input schema. See https://modelcontextprotocol.io/docs/concepts/tools for the correct pattern.
• Expand execute_skill_action description to 50-200 chars explaining what action it performs, when to call it, and what it returns. Example: 'Execute a skill action for flow matching latent diffusion synthesis. Accepts action_type and parameters. Returns execution status and result artifact URL.'
• Document output schemas for both tools. For synthesize_flux_diffusion_frame, declare that it returns an object with fields: generation_id (string), prompt (string), dimensions (string), flow_matching_euler_steps (integer), output_tensor_artifact_url (string), etc.
• Add input schema to execute_skill_action. Define what parameters it accepts (action_type, parameters, etc.), their types, and descriptions. Currently invisible.
• Implement error handling with recovery guidance. Return structured errors that classify failures as retryable vs. fatal, and suggest next steps. Example: 'Invalid resolution: width must be 512-2048. Try synthesize_flux_diffusion_frame with resolution_width=1024.'