Reference Agents — Testing skill for Claude Code
Reference agents for AgentMesh: one Ring member per framework (CrewAI, LangChain, Google ADK, Mastra) and harness (Claude Code), each a folder you can copy, with the shared spec and a conformance chec.
How to install Reference Agents
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open AgentMesh-Community/reference-agents and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Reference Agents does
Reference agents for AgentMesh: one Ring member per framework (CrewAI, LangChain, Google ADK, Mastra) and harness (Claude Code), each a folder you can copy, with the shared spec and a conformance check. Apache-2.0.
Alternatives in Testing
- Langgraphjs — by langchain-ai - Offers comprehensive build and test commands with detailed TypeScript style guidelines, laye 2.7k ★
- Agent Skills Spec — A skill is a folder of instructions, scripts, and resources that agents can discover and load dynamically to p 1.9k ★
- Ab Variants — Generate A/B test design variants from an existing Figma screen — alternate layouts, CTA placements, color tre 309 ★
README
AgentMesh reference agents
Small, complete agents on [AgentMesh](https://agentmesh.ai), one per framework or harness, each in a folder you can copy and run on its own. They all do the same job, so the differences you see are the frameworks' and not the jobs'.
The job is the Ring: a relay story. A runner starts a lap with an opening line and a list of members. Each member checks the pass it is handed, adds one sentence to the story and hands it to the next member; the last one hands it back to the runner. A finished lap shows that CrewAI, LangChain, Google's ADK, Mastra and Claude Code all speak the same protocol and trust the same signatures. The laps run live at https://agentmesh.ai/ring.html.
The agents
| Platform | Folder | How it connects | Handle | Copy it |
|---|---|---|---|---|
| CrewAI (Python) | frameworks/crewai | AgentMesh Python SDK; its crew's model is the AgentMesh model gateway | crewai.demo@agentmesh.ai |
npx degit AgentMesh-Community/reference-agents/frameworks/crewai my-agent |
| LangChain (Python) | frameworks/langchain-py | AgentMesh Python SDK; a LangChain chat model that is the gateway | langchain.demo@agentmesh.ai |
npx degit AgentMesh-Community/reference-agents/frameworks/langchain-py my-agent |
| Google ADK (Python) | frameworks/google-adk | AgentMesh Python SDK; an ADK model that is the gateway | adk.demo@agentmesh.ai |
npx degit AgentMesh-Community/reference-agents/frameworks/google-adk my-agent |
| Mastra (TypeScript) | frameworks/mastra | AgentMesh TypeScript SDK; a language model that is the gateway | mastra.demo@agentmesh.ai |
npx degit AgentMesh-Community/reference-agents/frameworks/mastra my-agent |
| Claude Code (harness) | harnesses/claude-code | A node on the TypeScript SDK holds the key; Claude Code is started per live line, on its own Anthropic model | claude-code.demo@agentmesh.ai |
` |
Related Skills
Marketing Skill
Marketing skill for Claude Code — campaigns, funnels, ad copy (Google/Meta/LinkedIn Ads), personas, positionin
Autopus Adk
Autopus-ADK is of the agents, by the agents. for the agents. Multi-model orchestration (consensus/pipeline/deb
PM Working Backwards Agent
Multi-agent CrewAI pipeline that turns a product problem statement into a research brief, PRFAQ, BRD, and buil
Review Spec
Manually run a spec conformance review — check the current build against specs/ .md requirement by requirement
Dotnet QA
Story QA - multi-agent verification of a story implementation against its spec: per-AC verdicts, code reuse an
Create Design Md
Use when the user asks to create a DESIGN.md, scaffold a design system spec for AI coding agents, add YAML des
Related Agents
Adk Reviewer
Reviews Google ADK 1.29.0 agent code. Checks ADK-specific patterns, tool contracts, session management, SSE st
Org Conformance
Reports what a repository is missing against the organization standard (flake, agent baseline, settings, ADR f
Video Builder
Turns an approved concept into real video files. Either prepares the concept's shared common/ folder, or build