This MCP server manages MCP servers and Skills within AstrBot, a Chinese-language bot framework. Tools are well-named with clear verbs (list_, get_, enable_, disable_, add_, update_, remove_, install_, delete_) and follow verb_noun conventions. However, quality varies significantly across tools. Most have descriptions (many in Chinese), but 4 tools (add_mcp_server, update_mcp_server, remove_mcp_server, delete_skill, install_skill, update_skill_from_zip) have minimal or generic English descriptions lacking detail about when to use them, what they return, and error recovery paths. All tools have basic input schemas with type definitions, but parameter descriptions are sparse or missing for several tools (e.g., 'args' in add_mcp_server lacks guidance on format; 'force' in install_skill lacks explanation of what security issues it overrides). Output schemas are entirely undocumented, callers cannot see what fields to expect from responses. Error handling is present in code (permission checks, validation) but descriptions do not explain error cases or recovery paths. No tool has destructive/readonly/idempotent hints. Tool composition is reasonable, separate tools for enable/disable, but install/update could benefit from clearer error messaging about what 'force' actually does.

• Document output schemas for all tools in MCP tool definitions. For example, list_mcp_servers should declare: { type: object, properties: { servers: { type: array, items: { type: object, properties: { name, active, status, tools, transport } } } } }. This enables LLMs to parse responses and chain tool calls.
• Expand descriptions for add_mcp_server, update_mcp_server, remove_mcp_server, delete_skill, install_skill, and update_skill_from_zip to 80 - 150 chars. Include: what action is taken, when to use it instead of similar tools, what permissions are required, what the tool returns, and what can go wrong. Example: 'Add a new MCP server to the configuration. Requires admin. The server will be loaded on next restart unless enabled immediately. Returns { ok: true, server_name: ... } or { ok: false, error: ... }. Name must match [A-Za-z0-9._-]+.'
• Add detailed parameter descriptions for: 'args' (format: JSON array of strings, e.g., ['--port', '5000']), 'env' (key-value pairs; no secrets), 'force' (bypass security scan warnings; use only if you trust the source), 'zip_path' (local file path or HTTP URL to a .zip file), 'skill_name' (alphanumeric + underscore; used in commands).
• Add error recovery guidance to descriptions. For example, enable_mcp_server: 'If the server does not exist, use add_mcp_server first. If connection fails, check the command/URL and try again. If you lack admin permission, ask the owner.'
• Annotate tool definitions with readOnlyHint (list_*, get_*), destructiveHint (remove_*, delete_*), and idempotentHint (enable_*, disable_*) for safer composition by MCP clients.