agent-init

Generates agent.yaml configuration and .workflow scaffolding for Claude Agent projects.

Updated May 2, 2026
One-click install
npx skills add https://github.com/whinchman/midi-man-mk3 --skill agent-init-whinchman
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: agent-init
Source: https://github.com/whinchman/midi-man-mk3/tree/main/.claude/skills/agent-init
Command: npx skills add https://github.com/whinchman/midi-man-mk3 --skill agent-init-whinchman

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Setting up the Claude Agent framework in a new or existing project requires manually creating configuration files, workflow directories, and gitignore rules, which is error-prone and inconsistent across projects. ## Core Features & Use Cases - Interactive Setup Wizard: Walks through project identity, stack detection, testing, build commands, and code standards step by step, confirming everything before writing files. - Single-Repo and Multi-Repo Modes: Detects nested git repositories and configures either a single agent.yaml or a root config with per-repo agent-build.yaml files, including mono-repo sub-project support. - Workflow Backend Selection: Scaffolds either local markdown tracking files (BACKLOG.md, TODO.md, DONE.md, BUGS.md) or provisions a GitHub Project board via board-man scripts. - Use Case: A developer clones a Rust project and runs /agent-init; the wizard detects Cargo.toml, proposes cargo test defaults, and writes a complete agent.yaml plus .workflow/ scaffold ready for the coordinator pipeline. ## Quick Start Run /agent-init in your project root and answer the wizard's questions to generate the agent configuration and workflow files.

Frequently Asked Questions about agent-init

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

FAQPage Schema
How do I set up the Claude Agent framework in an existing project?▼

Run /agent-init in the project root. The wizard detects your stack from manifest files like Cargo.toml or package.json, asks for project identity and test commands, then writes agent.yaml and the .workflow/ scaffold after your confirmation.

How does agent-init handle monorepos with multiple git repositories?▼

It searches subdirectories for .git folders and offers multi-repo mode. Each leaf repo gets a slim agent-build.yaml with its own stack and build config, while the root agent.yaml holds shared project, git, and workflow settings.

What is the difference between the markdown and github_project workflow backends?▼

The markdown backend tracks work in local .workflow files like BACKLOG.md and TODO.md. The github_project backend uses a GitHub Project board with FEATURE, CHANGE, and BUG issues, provisioned by the board-man-setup.sh script, and skips the markdown files.

Does agent-init overwrite an existing agent.yaml file?▼

No, it warns that agent.yaml already exists and asks for explicit confirmation before overwriting. It also shows a full YAML preview of the configuration before writing any files.

What happens if my project is not a git repository?▼

The wizard still completes setup but warns that no git repository was detected. You must run git init before using the agent workflow, since worktrees and branching require git.