Kiro Crew Project Setup Skill

A reusable Agent Skill for designing, hardening, validating, and troubleshooting a Kiro Crew setup for a software or content project.

The skill helps an AI coding assistant turn project requirements into:

• A project operating guide with explicit authority and human approval gates.
• Narrowly scoped coordinator, developer, editorial, and GitHub issue agents.
• Safe GitHub Issue Radar automation that starts with one opt-in issue at a time.
• Isolated worktree and draft-pull-request workflows.
• Gateway and system-service diagnostics.
• Validation and smoke-test evidence before unattended work is enabled.

The package follows the open Agent Skills directory format and can be used by Codex, Claude Code, and other compatible assistants.

Install

Clone the repository:

git clone https://github.com/koushal2018/kirocrew-project-setup-skill.git
cd kirocrew-project-setup-skill

Codex

Install for your user:

mkdir -p ~/.agents/skills
cp -R skills/kirocrew-project-setup ~/.agents/skills/

Or install only for one repository:

mkdir -p /path/to/project/.agents/skills
cp -R skills/kirocrew-project-setup /path/to/project/.agents/skills/

Invoke it with:

$kirocrew-project-setup

Claude Code

Install for your user:

mkdir -p ~/.claude/skills
cp -R skills/kirocrew-project-setup ~/.claude/skills/

Or install only for one repository:

mkdir -p /path/to/project/.claude/skills
cp -R skills/kirocrew-project-setup /path/to/project/.claude/skills/

Invoke it with:

/kirocrew-project-setup

Other assistants

Copy the complete skills/kirocrew-project-setup directory into the assistant's documented Agent Skills location. Keep SKILL.md, assets/, references/, and scripts/ together so relative links continue to work.

Recommended first prompt

Use the kirocrew-project-setup skill to inspect this repository and propose a safe Kiro Crew team. Do read-only discovery first. Do not install a service, change GitHub state, enable unattended work, publish, deploy, or merge until I explicitly approve those actions.

What is included

skills/kirocrew-project-setup/
├── SKILL.md
├── agents/openai.yaml
├── assets/
│   ├── project-operating-guide.template.md
│   └── agents/
│       ├── coordinator.template.json
│       ├── developer.template.json
│       └── github-issue-developer.template.json
├── references/
│   ├── github-issue-radar.md
│   ├── installation.md
│   ├── service-troubleshooting.md
│   └── team-design.md
└── scripts/kirocrew_preflight.py

The preflight script is read-only. It checks the local toolchain, repository state, custom-agent validation, delegation settings, gateway reachability, and GitHub CLI context without installing services or changing GitHub:

python3 skills/kirocrew-project-setup/scripts/kirocrew_preflight.py --repo /path/to/project

Add --json for machine-readable output or --with-doctor to include the slower kirocrew doctor check.

Safety model

The default workflow intentionally keeps these actions behind explicit human approval:

• Enabling unattended issue processing.
• Creating or modifying GitHub labels and repository settings.
• Marking pull requests ready or merging them.
• Deploying applications or infrastructure.
• Publishing articles or sending external messages.
• Reading secrets, force-pushing, destructive cleanup, or weakening CI gates.

GitHub issue automation uses a dedicated opt-in label, a starting concurrency of one, isolated worktrees, draft pull requests, and multiple independent safeguards. Adapt every template to the repository and validate it with the installed Kiro CLI before use.

Compatibility

The skill inspects installed command help rather than assuming a fixed CLI. Its templates and checks were validated while this repository was created with:

• Kiro Crew 0.5.0rc1
• Kiro CLI 2.20.1

Kiro Crew is evolving. Treat version-specific notes as diagnostic leads and verify behavior in the actual gateway runtime.

Contributing

Issues and pull requests are welcome. Please read CONTRIBUTING.md before submitting changes.

License and trademarks

Released under the MIT License.

This is an unofficial community project. It is not affiliated with or endorsed by Kiro, Kiro Crew, AWS, OpenAI, or Anthropic. Product names and trademarks belong to their respective owners.