Agent project scaffold

Skill minagayid/Agents/skills/agent-project-scaffold

Scaffold a new AI agent project using a chosen SDK (Claude Agent SDK, OpenAI Agents SDK, or Google ADK). Use when the user asks to "start", "bootstrap", or "scaffold" an agent, or to set up a new agent codebase with a tool and an MCP server wired in.From its SKILL.md

Install

npx -y skills add minagayid/Agents --skill agent-project-scaffold

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.3 KB, as published. Nobody here has run it

Agent Project Scaffold

A reusable Agent Skill (following the Agent Skills standard) that guides an agent through bootstrapping a new AI‑agent project from a clean, runnable starting point.

When to use

Trigger when the user wants to create a new agent project and hasn't specified every detail — e.g. "set up a Claude agent with a weather tool", "scaffold an OpenAI agents project", or "bootstrap a Google ADK agent".

Steps

Pick the SDK. If unstated, ask: Claude Agent SDK (Python/TS), OpenAI Agents SDK (Python), or Google ADK (Python). Default to Claude Agent SDK Python.
Copy the starter. Use the matching folder in this repo's sdk/ directory as the template (quickstart.py/.ts, plus requirements.txt/package.json).
Set the env var. ANTHROPIC_API_KEY, OPENAI_API_KEY, or GOOGLE_API_KEY respectively.
Add a tool. Define one function tool relevant to the user's domain, using the SDK's tool pattern (@tool / @function_tool / a plain function passed in tools=[...]).
(Optional) Wire an MCP server. Add an entry from mcp/mcp-servers.json (e.g. filesystem or fetch) so the agent can reach external tools.
Install & run. Install deps and run the quickstart to confirm it works end‑to‑end.
Confirm. Report the run output and the file layout you created.

Guidelines

Keep the first version minimal and runnable; add complexity only after it runs.
Never hard‑code API keys — read them from environment variables.
Prefer the SDK's async entrypoint (Runner.run, run_async, async for … query()).
Match the surrounding project's language and conventions if scaffolding inside an existing repo.

Examples

"Bootstrap a Claude agent that can read files." → Claude Agent SDK Python + filesystem MCP server.
"Scaffold an OpenAI agent with a calculator tool." → OpenAI Agents SDK + @function_tool.
"Start a Google ADK weather agent." → ADK Python root_agent + adk web.