cuyamaca-runtime-agent

Run a runtime agent loop issuing serial commands from sensor and camera data.

Updated Mar 31, 2026
One-click install
npx skills add https://github.com/yuyanghu06/cuyamaca --skill cuyamaca-runtime-agent
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: cuyamaca-runtime-agent
Source: https://github.com/yuyanghu06/cuyamaca/tree/main/.claude/skills/cuyamaca-runtime-agent
Command: npx skills add https://github.com/yuyanghu06/cuyamaca --skill cuyamaca-runtime-agent

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill helps developers build and integrate a runtime agent loop for Cuyamaca, enabling a runtime model to observe sensor data, camera frames, and multimodal context, decide which tools to call, issue serial commands, and iterate until the user stops it.

Core Features & Use Cases

  • Runtime agent loop: observe, decide, act, and repeat with a kill-safe flow.
  • Multimodal context assembly: combine sensor state, sensor visualizations, and camera frames into model prompts.
  • Tool call dispatch via serial: translate model tool calls into CMD serial commands to the board.
  • Kill button and safety: always-on emergency stop and controlled termination of sessions.
  • Phase-agnostic references: applicable to phases 6 and 7 workflows in Cuyamaca.

Quick Start

Start the runtime agent by connecting a flashed board, initialize multimodal context, and begin the observe–decide–act loop via tool calls.

Frequently Asked Questions about cuyamaca-runtime-agent

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I build a runtime agent loop to control an Arduino robot via serial commands?

A runtime agent loop controls an Arduino robot by continuously observing sensor data and camera frames, deciding which tools to call, and issuing serial commands to the board until the user stops it.

How does multimodal context assembly work for robot control?

Multimodal context assembly combines current sensor states, sensor visualizations, and live camera frames into model prompts, enabling the runtime agent to make informed decisions about which serial commands to issue next.

What prerequisites do I need before starting a runtime agent for serial communication?

Starting a runtime agent requires Phase 6 readiness, including established serial communication, sensor parsing, and sensor visualization, along with a configured runtime model slot and a flashed board.

How do I handle safe termination and emergency stops during a robot runtime agent loop?

Safe termination is handled via an always-on kill button for emergency stops and a controlled end_session process, ensuring the runtime agent and serial communication stop immediately when triggered.

How are model tool calls translated into serial commands for the board?

Model tool calls are dispatched directly as CMD serial commands to the connected board, translating the runtime agent's multimodal decisions into physical robot actions during the control loop.