Add server or client AI chat tools to the ryOS assistant using Zod schemas and handlers.
npx skills add https://github.com/ryokun6/ryos --skill add-ai-chat-tool
Please help me install this Agent Skill. Skill: add-ai-chat-tool Source: https://github.com/ryokun6/ryos/tree/main/.cursor/skills/add-ai-chat-tool Command: npx skills add https://github.com/ryokun6/ryos --skill add-ai-chat-tool
Adding a new capability to the ryOS AI assistant requires touching many files across the server and client, and it is easy to miss a step such as the dispatch switch case or the execution-side metadata. This Skill provides the complete file map, decision rules, and code patterns so a new chat tool works end to end on the first attempt.
execute function (Redis, secrets, server fetch) or a client-side handler (Zustand stores, IndexedDB, DOM, media playback).Ask the AI to add a new chat tool named myFeatureControl following the add-ai-chat-tool skill, with list, create, and delete actions handled client-side.
High-intent search queries and answers about installing and using this skill.
Define a Zod schema in api/chat/tools/schemas.ts, add a description and tool entry in api/chat/tools/index.ts, then wire either a server executor or a client handler in src/apps/chats/tools with a dispatch switch case. The Skill provides code templates for each step.
Use server execution when the tool needs Redis, secrets, or server-side fetch, and client execution when it touches Zustand stores, IndexedDB, media playback, app windows, or the DOM. Some tools like stickiesControl run both ways for web chat and Telegram profiles.
The client dispatch is an explicit switch statement with no automatic registry. You must add a case for your tool name in the dispatch file, otherwise the call falls through to the default branch and reports an unhandled tool error.
Add a schema unit test that safeParses valid and invalid inputs under tests/unit, cover server executors through the AI endpoint suite, and verify client handlers by asking Ryo in the Chats app and confirming the store updates and tool result bubble.
Yes, server executors can return images by adding a toModelOutput function that converts the result into multimodal content, following the pattern used by infiniteMacControl's readScreen capability.
silvarojasdiegoarturo91Guide Shopify app developers through AI integration, MCP updates, and schema validation.
AgentientValidate Firestore document reads and writes with Zod schemas via withConverter.
berkcangumusisikScan React, Vue, and Angular files to replace hardcoded strings with i18n keys.
UsernameTronGenerate Claude Code SKILL.md files from natural language descriptions.