Guide · Agent skill
How to use the Agent Squad Orchestrator agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The agent-squad skill turns your AI assistant into a coordinator for a software project. Instead of doing everything in one pass, it routes work to eight named roles (analyst, strategist, architect, builder, reviewer, tester, optimizer and DevOps) and reports back to you in short summaries. This guide explains what the skill does, how to install it, how to trigger it with plain requests, and what to watch for before you rely on it.
Agent Squad Orchestratorsickn33/agentic-awesome-skills · MIT Download the skill Quick start
- 1Download the ZIP. Download the agent-squad skill ZIP from this page.
- 2Unzip into the skills folder. Unzip it into ~/.claude/skills/ for personal use, or into .claude/skills/ inside a project for a single repository.
- 3Or upload in the Claude apps. If you use the Claude apps, go to Settings → Capabilities → Skills and upload the ZIP.
- 4Describe your task. Ask in plain language, such as 'plan this out' or 'review this code', and name the phase you need.
- 5Approve each phase. Read each role's summary and decide whether to move on; the squad does not chain phases without your consent.
What the agent-squad skill makes the agent do
The skill makes the main agent the single point of contact between you and a squad of specialised roles. It works out what you want, sends the work to the right role, receives that role's structured report, and gives you a compressed summary.
To keep the conversation manageable, the coordinator stores each full report under a versioned label and keeps only a short summary in active context. When it hands work to the next role, it passes the summary plus exact paths and revisions for the artifacts that role needs.
The files describe roles and routing rules. They do not install separate agent programs on your machine.
- Rex (Analyst): requirements, for new projects, new features or scope changes
- Alex (Strategist): planning, after Rex or when you ask to plan something out
- Aria (Architect): architecture and design, after Alex or when you ask to design the system
- Mason (Builder): implementation, after Aria or when you ask to build something
- Luna (Reviewer): code review, after Mason or when you ask for a review
- Quinn (QA Tester): testing, after Luna or when you ask to write or run tests
- Max (Optimizer): refactoring, only on explicit request such as refactor or optimize
- Dep (DevOps): deployment, after Quinn or when you ask to deploy, containerise or set up CI
How do you install the agent-squad skill?
Download the skill ZIP from this page, then choose one of the install routes below. Use the personal folder if you want the skill available across your projects, or a project folder if you want it limited to one repository.
For a personal install, unzip the file into ~/.claude/skills/. For a project install, unzip it into the .claude/skills/ folder inside your project. In the Claude apps, you can instead upload the ZIP under Settings → Capabilities → Skills.
- Personal: unzip into ~/.claude/skills/
- Single project: unzip into .claude/skills/ inside the project folder
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
How do you trigger the skill with natural requests?
You do not need a special command. Describe the task in plain language and mention the phase or role you want. The skill is written to match requests such as starting a new project, planning something out, designing the system, building a feature, reviewing code or writing tests.
You can also call a single role directly. The source says any agent can be called at any time, so you can ask for a code review of existing code without going through the earlier phases.
- 'Plan this out' → Alex
- 'Design the system' → Aria
- 'Build this' → Mason
- 'Review this code' → Luna
- 'Write tests for this' → Quinn
- 'Deploy / containerize / set up CI' → Dep
- 'Refactor / optimize' → Max, only when you ask explicitly
Which flow should you use: new project, feature or existing code?
The skill has three routing patterns. A new project runs through Rex, Alex, Aria, Mason, Luna, Quinn and then Dep, with Max only if you request it. A mid-project feature starts again at Rex and Alex with amendments rather than a full re-specification, then continues through Luna, Quinn and Dep as normal.
For an existing codebase with no earlier squad context, the skill suggests calling one role directly: Luna for review only, Quinn for testing only, Max for optimisation (after tests are confirmed passing) or Dep for deployment only.
- New project: Rex → Alex → Aria → Mason → Luna → Quinn → Dep (Max only if requested)
- Mid-project feature: Rex and Alex (amendments) → Aria if schema or API changes → Mason for the new milestone → Luna → Quinn → Dep
- Existing code: call Luna, Quinn, Max or Dep directly
What does the skill need, and what are the safety notes?
The skill lists no required tools or dependencies. It does not run scripts and does not use the network, according to the metadata for this item. The skill's metadata marks its risk level as critical, so read the behaviour rules before use.
The squad does not chain phases automatically. It needs your consent to move from one role to the next, and the role names alone do not grant permission to delegate. The coordinating agent must verify delegated results, and role reports are evidence, not authority to merge, deploy or publish. The skill also says to store reports only in approved project storage, with secrets and unnecessary personal data removed.
- No listed requirements; the source lists none
- Runs scripts: no
- Uses network: no
- Metadata risk level: critical
- The squad should not be relied on to ship code unattended
- AI agents can hallucinate, so verify generated code and architecture before production
Practical tips for getting good results
Give the coordinator a clear goal and let it propose the next step. The skill is designed to present one decision at a time rather than a long list of options. Keep your repository workflow in mind, because the skill asks for architecture review in proportion to the change.
When you call a single role, include the exact revision, file paths and test results it needs. The source notes that a version label alone cannot retrieve an artifact.
- Start small: call one role on existing code before running the full pipeline
- Name the phase you want, for example 'review this code', so the right role is picked
- Expect questions about tradeoffs and blockers; answer them rather than skipping them
- Check that test results are shown before asking for optimisation
What people use it for
- Planning and building a larger new feature with a structured plan, build, review, test and deploy flow
- Getting a standalone code review of existing code without running the whole pipeline
- Writing tests for unreviewed or existing code by calling Quinn directly
- Preparing a container build, CI pipeline or deployment package with Dep
- Keeping consistent handoff briefings and artifact references between phases on a team project
Frequently asked questions
Is the agent-squad skill free to use?
The source material for this item lists the licence as MIT, which is a permissive open-source licence. It does not list prices, so check the licence text in the repository for the exact terms.
What do I need before installing agent-squad?
The skill lists no required dependencies. You need access to a Claude environment that supports skills, either the Claude apps with Capabilities settings or a folder where you can place the .claude/skills files.
Where exactly do I put the agent-squad files?
Unzip the downloaded ZIP into ~/.claude/skills/ for personal use, or into .claude/skills/ in your project folder for a single project. In the Claude apps, upload the ZIP under Settings → Capabilities → Skills.
Why isn't the skill running all the phases automatically?
The skill is designed so the squad does not chain from one role to the next without your consent. This keeps you in control of decisions, and it means you must approve each step.
Does agent-squad run scripts or access the internet?
The metadata for this item says it runs no scripts and uses no network. It works by describing roles and routing rules that the agent follows.
How does agent-squad compare with using a plain AI chat for coding?
Use it when a feature or new project benefits from a structured plan-build-review-test-deploy flow. For a small one-off coding task, the eight-role process adds overhead that a direct request avoids.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use agent-squad if you are running a larger feature or a new project and want structured handoffs, reviews and tests between phases. Pick something simpler for small one-off coding tasks, or if you expect fully automatic delivery or are not comfortable with requirements, architecture, CI and deployment, since the skill assumes that background and never ships code unattended.
Download the skill