How do I install the Ryu Tools skill?

Run npx skillmds@latest add amajorai/ryu-tools in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Ryu Tools safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Ryu Tools?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Ryu Tools free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Ryu Tools?

amajorai (@amajorai) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Ryu Tools by amajorai — Agent Skill

Name: Ryu Tools
Author: amajorai

Ryu Tools

Reference for the ryu_* MCP tools this plugin exposes and how to combine them to drive a Ryu node. Covers node health, models and engines, agents and teams, skills, workflows, the MCP bridge, RAG retrieval, and one-shot ask. Use when deciding which Ryu tool to call or how to chain them for a task.

amajorai Updated Sep 21, 2026

File contents

Driving Ryu through the ryu_* tools

This plugin bundles an MCP bridge to one Ryu Core node. Every tool below is namespaced ryu_ and maps to a typed Core API call. tools/list reflects the live set - trust it for exact names and argument shapes over this summary. All tools are read-only except ryu_set_active_model, ryu_install_skill, and a normal ryu_run_workflow call. Pass dryRun: true to ryu_run_workflow for a transient read-only preview that creates no run history and skips effectful nodes.

Health and system

ryu_health - is the node alive. Call this first when anything fails.
ryu_system_status - active engine, engine running state, sidecars, gateway reachability, mesh.
ryu_system_info - hardware snapshot: CPU, RAM, disk, GPU/VRAM, OS.

Models and engines

ryu_search_models { query, limit? } - search the catalog (Hugging Face GGUF by default).
ryu_get_active_model - what the local chat stack is serving now.
ryu_set_active_model { modelId, engine? } - switch the served model. It must already be installed; the engine is usually derived from the model format.
ryu_list_engines - inference engines and their installed models.

Agents and teams

ryu_list_agents - agents configured on the node.
ryu_list_teams - multi-agent teams.

agentId values from ryu_list_agents are what ryu_call_mcp_tool needs, because Core ties each MCP tool allowlist to a registered agent.

Skills

ryu_list_skills - installed skills and their active state.
ryu_search_skills { query, limit? } - search the skills directory.
ryu_install_skill { id } - install a skill by catalog id and hot-reload Core's registry.

Workflows

ryu_list_workflows - defined workflows.
ryu_run_workflow { id, input?, dryRun? } - run one by id. dryRun: true is read-only and transient; a normal run can return awaiting_input, which should be surfaced exactly.

MCP bridge (reach any registered server)

ryu_list_mcp_servers - MCP servers Ryu has registered.
ryu_call_mcp_tool { tool, server?, agentId, args? } - invoke a tool on any registered MCP server. tool may be fully qualified server.tool or a bare name plus server. agentId is required. This is how you reach Ghost, Shadow, Composio, and every other server Ryu governs, through one call.

Knowledge and memory (RAG)

ryu_list_spaces - knowledge Spaces (document collections).
ryu_search_space { spaceId, query, limit? } - semantic search inside one Space.
ryu_search_retrieval { query, topK? } - unified search over memory plus all Spaces.

One-shot ask

ryu_ask { question, conversationId? } - ask Ryu's side-question model one question with no tool access. Fast and self-contained. Not an agent run; for that, inspect ryu_list_agents or run a workflow.

Common chains

Bring a node online: ryu_health then ryu_get_active_model, and if empty, ryu_search_models then ryu_set_active_model.
Answer from the user's own knowledge: ryu_search_retrieval, then cite the hits.
Use a governed external tool: ryu_list_agents for an agentId, ryu_list_mcp_servers for the server, then ryu_call_mcp_tool.

See [[setup-ryu]] to get a node running in the first place.

---name: ryu-toolsdescription: Reference for the ryu_* MCP tools this plugin exposes and how to combine them to drive a Ryu node. Covers node health, models and engines, agents and teams, skills, workflows, the MCP bridge, RAG retrieval, and one-shot ask. Use when deciding which Ryu tool to call or how to chain them for a task.---# Driving Ryu through the ryu_* toolsThis plugin bundles an MCP bridge to one Ryu Core node. Every tool below is namespaced `ryu_` and maps to a typed Core API call. `tools/list` reflects the live set - trust it for exact names and argument shapes over this summary. All tools are read-only except `ryu_set_active_model`, `ryu_install_skill`, and a normal `ryu_run_workflow` call. Pass `dryRun: true` to `ryu_run_workflow` for a transient read-only preview that creates no run history and skips effectful nodes.## Health and system- `ryu_health` - is the node alive. Call this first when anything fails.- `ryu_system_status` - active engine, engine running state, sidecars, gateway reachability, mesh.- `ryu_system_info` - hardware snapshot: CPU, RAM, disk, GPU/VRAM, OS.## Models and engines- `ryu_search_models` `{ query, limit? }` - search the catalog (Hugging Face GGUF by default).- `ryu_get_active_model` - what the local chat stack is serving now.- `ryu_set_active_model` `{ modelId, engine? }` - switch the served model. It must already be installed; the engine is usually derived from the model format.- `ryu_list_engines` - inference engines and their installed models.## Agents and teams- `ryu_list_agents` - agents configured on the node.- `ryu_list_teams` - multi-agent teams.`agentId` values from `ryu_list_agents` are what `ryu_call_mcp_tool` needs, because Core ties each MCP tool allowlist to a registered agent.## Skills- `ryu_list_skills` - installed skills and their active state.- `ryu_search_skills` `{ query, limit? }` - search the skills directory.- `ryu_install_skill` `{ id }` - install a skill by catalog id and hot-reload Core's registry.## Workflows- `ryu_list_workflows` - defined workflows.- `ryu_run_workflow` `{ id, input?, dryRun? }` - run one by id. `dryRun: true` is read-only and transient; a normal run can return `awaiting_input`, which should be surfaced exactly.## MCP bridge (reach any registered server)- `ryu_list_mcp_servers` - MCP servers Ryu has registered.- `ryu_call_mcp_tool` `{ tool, server?, agentId, args? }` - invoke a tool on any registered MCP server. `tool` may be fully qualified `server.tool` or a bare name plus `server`. `agentId` is required. This is how you reach Ghost, Shadow, Composio, and every other server Ryu governs, through one call.## Knowledge and memory (RAG)- `ryu_list_spaces` - knowledge Spaces (document collections).- `ryu_search_space` `{ spaceId, query, limit? }` - semantic search inside one Space.- `ryu_search_retrieval` `{ query, topK? }` - unified search over memory plus all Spaces.## One-shot ask- `ryu_ask` `{ question, conversationId? }` - ask Ryu's side-question model one question with no tool access. Fast and self-contained. Not an agent run; for that, inspect `ryu_list_agents` or run a workflow.## Common chains- Bring a node online: `ryu_health` then `ryu_get_active_model`, and if empty, `ryu_search_models` then `ryu_set_active_model`.- Answer from the user's own knowledge: `ryu_search_retrieval`, then cite the hits.- Use a governed external tool: `ryu_list_agents` for an `agentId`, `ryu_list_mcp_servers` for the server, then `ryu_call_mcp_tool`.See [[setup-ryu]] to get a node running in the first place.

Frequently asked questions