crilsen

AI Agent Relay — AI skill for Claude Code

AI community

Portable context for AI coding agents and agentic harnesses.

How to install AI Agent Relay

This entry records only its repository, not the path inside it, so there is no exact command to give. Open crilsen/ai-agent-relay and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What AI Agent Relay does

Portable context for AI coding agents and agentic harnesses.

Alternatives in AI

DeepLearning.Ai Courses ClaudeCode — Vibe Genius 团队整理的 DeepLearning.AI 官方课程 'Claude Code: A Highly Agentic Coding Assistant' 学习笔记，由 Andrew Ng 和 Ant 326 ★
Relay Radar — AI中转站质量监控 · 模型验真 · Token成本优化 — RelayRadar 中转雷达 131 ★
Gensee Crate — Run AI harnesses like Claude Code, Codex, Copilot, Antigravity, and Omnigent in disposable workspace forks 123 ★

README

ai-agent-relay

[![lint](https://github.com/crilsen/ai-agent-relay/actions/workflows/lint.yml/badge.svg)](https://github.com/crilsen/ai-agent-relay/actions/workflows/lint.yml) [![license: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE) [![context: AGENTS.md](https://img.shields.io/badge/context-AGENTS.md-2ea44f.svg)](AGENTS.md)

A portable, Git-versioned context layer for software projects that use coding agents, agentic harnesses, or multiple AI tools over time. Pass the baton between agents, models, providers, and machines without losing context.

Keep project knowledge in the repository instead of a chat history, so any agent can resume work regardless of model, IDE, provider, or computer. Works with Codex, Cursor, Kiro, OpenCode, Claude Code, DeepSeek Harness, Cline, Roo Code, and future tools.

**Use at your own risk.** Provided as is, without warranty. You are responsible for reviewing context before an agent acts on it, for the agent's actions, and for all costs including AI provider usage and token consumption. See [SECURITY.md](SECURITY.md).

Harness-ready

This template is ready to use in projects built with agentic harnesses. The portable core — `AGENTS.md`, `.ai/`, per-tool adapters, the Resume block, checkpoint/limits, and the switch-agent protocol — applies as-is to a harness project, so an agent running under any harness can pick up the context without extra setup.

It composes with harness features instead of replacing them: hooks trigger the checkpoint and handoff, subagents read the same context, skills can live under `.ai/skills/`, and MCP tools are registered in `.ai/TOOLS.md`. See [docs/harness-integration.md](docs/harness-integration.md).

Note what it is: the template is the portable context layer that a harness reads, not a harness itself. It carries no runtime.

The bundled infrastructure workflows (`terraform-change`, `kubernetes-change`, `cloud-port`) are optional; a harness or application project can ignore or rem