This is a VS Code extension, not an MCP server. The source shows only package.json command definitions with minimal descriptions (10-20 chars each) and NO visible input schemas, parameter types, or output documentation. Tools are registered as VS Code commands, not MCP tools with proper schemas. No evidence of MCP protocol implementation in the provided code. Tool descriptions like 'Install Skill', 'Search Skills', 'Doctor' are far too generic and lack context for LLM selection. No parameter documentation visible. This cannot be evaluated as an MCP server.

• Expand all tool descriptions to 50-200 characters. Example: 'rolecraft.install' → 'Install a skill from the registry to one or more AI agents (Cursor, Claude Code, etc). Requires skill name and target agent(s).'
• Define explicit input schemas for all tools with JSON Schema. Specify parameter types (string, array, enum), required fields, and constraints. Example: rolecraft.search should accept query (string, required), limit (integer, 1-100, default 20), offset (integer, default 0).
• Document output schemas. Example: rolecraft.list should return {skills: [{name, version, description, installed_agents}], total_count, next_cursor}.
• Add confirmation/dry-run support for destructive tools. Implement a 'confirm' parameter or separate 'preview_delete' tool to prevent accidental data loss.
• Add parameter descriptions for all inputs. Example: rolecraft.search query param → 'Search term to find skills by name or keyword. Supports partial matches.'