Guide · Agent skill
How to use the Antigravity Workflows agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
Antigravity Workflows is an agent skill that takes a complex goal, such as shipping a SaaS MVP or running a web security audit, and turns it into an ordered sequence of steps. Each step is tied to a recommended skill, an expected output and a check that must pass before the agent moves on. This guide explains what the skill does, how to install it, how to trigger it with normal requests and what to review before you rely on it.
Antigravity Workflowssickn33/agentic-awesome-skills · MIT Download the skill Quick start
- 1Download and unzip the skill. Download the ZIP from the directory page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder. In the Claude apps, you can instead upload the ZIP under Settings, then Capabilities, then Skills.
- 2Install the companion skills you need. Check that the referenced skills for your chosen workflow, such as ship-saas-mvp or security-audit-web-app, are present locally. Missing skills will be reported as gaps.
- 3Preview any skill installs first. Before installing other skills, review their IDs and support files and run the direct installer's dry run with the selected IDs and destination. Only install within your authorization.
- 4Make a request in plain language. Describe your outcome, for example shipping an MVP or running a security audit, and mention the constraints and any credentials or infrastructure the work needs.
- 5Review the checkpoints and final report. Watch each step's expected artifact and check. At the end, read the completed artifacts, validation evidence and remaining risks, then decide the next actions.
What does the antigravity-workflows skill make the agent do?
The skill is for goals that span several phases, such as plan, build, test and ship. Instead of you choosing each relevant skill by hand, the agent proposes the one or two workflows that fit your outcome and then works through the steps in order.
For every step, the agent announces what it is doing and what artifact it should produce, invokes the recommended skills, and verifies completion before continuing. At the end it reports completed artifacts, validation evidence and remaining risks with next actions.
The skill coordinates other skills. It does not do the specialised work itself, so the quality of each step depends on the companion skills installed on your machine.
- Turns a complex objective into a guided sequence of skill invocations
- Suggests which listed workflow fits best
- Keeps a check on each step and retains failures until they are fixed
- Reports artifacts, validation evidence and open risks at the end
How do you install the Antigravity Workflows skill?
Download the skill ZIP from the directory page, then unzip it into one of two locations. Use ~/.claude/skills/ to make it available to you across projects, or use the .claude/skills/ folder inside a specific project to limit it to that project.
If you use the Claude apps instead of the file system, you can upload the ZIP under Settings, then Capabilities, then Skills.
The skill's own rules say installing other skills must be reviewed first. Its installer documentation refers to a direct installer with a --dry-run option for previewing selected skill IDs before anything is installed. The reviewed files do not fully specify that installer, so check its exact name and usage in your own environment before running it.
- Option A: unzip into ~/.claude/skills/ for personal use
- Option B: unzip into a project's .claude/skills/ folder for one project
- Option C: upload the ZIP in the Claude apps under Settings, then Capabilities, then Skills
~/.claude/skills/How do you trigger the skill with a natural request?
You do not need special syntax. Name the skill and describe the outcome you want, or name one of the standard workflows. The skill's description matches requests to ship a SaaS MVP, audit application security, build an AI agent, run browser QA or design a domain model.
The default routing maps each kind of request to a workflow: product delivery goes to ship-saas-mvp, a security review to security-audit-web-app, an agent or LLM product to build-ai-agent-system, end-to-end or browser testing to qa-browser-automation, and domain-driven design to design-ddd-core-domain.
The examples below are the copy-paste prompts from the skill's own documentation. Replace the project details with your own.
Use @antigravity-workflows to run the "Ship a SaaS MVP" workflow for my project idea.What do you need before the skill is useful?
The skill depends on other skills being present locally. The workflows reference companion skills such as ship-saas-mvp, security-audit-web-app and qa-browser-automation. If a required skill is missing, the agent is meant to state the gap and fall back to the closest available skill.
You also need any project credentials or infrastructure the chosen workflow needs. The skill does not guarantee success without environment access, credentials or required infrastructure. If a prerequisite is unavailable, the agent should report the exact blocked step and continue with independent work.
For stack-specific browser automation in Go, the go-playwright skill may need its own corresponding skill present in your local skills repository.
- Companion skills such as ship-saas-mvp, security-audit-web-app and qa-browser-automation installed locally
- Project credentials and infrastructure that the chosen workflow needs
- Your project's existing commands and conventions, which the workflow cards tell the agent to follow
What are the safety notes for scripts and network use?
The skill's listed requirements say it runs no scripts and uses no network access. Its own guardrails say destructive actions need explicit user approval, and security testing requires explicit authorization.
Installation is the sensitive part. The skill says the exact skill IDs and support files should be reviewed before installing, using a dry run with the selected IDs and destination. Installs happen only within your authorization, and a failed preview never authorizes installation. Deployment is also only to be attempted when it is authorized.
- Runs scripts: no
- Uses network: no
- Review selected skill IDs and support files with a dry run before installing
- Get explicit authorization before any security testing
- Deployment is attempted only when authorized
What tips help you get the best results?
Start with a concrete outcome. The skill works best when you say what you want to end with, such as a working journey with a passing test, rather than a broad topic. Ask for a single workflow if you already know which one fits.
Expect checkpoints. When a check fails, the agent keeps the failure, fixes the input or implementation, reruns that check and continues only after it passes. Let it do that instead of skipping steps.
Keep it for multi-phase work. For a small edit or a simple question, the orchestration adds overhead without benefit.
- Name the outcome and the constraints in the first message
- Confirm which companion skills are installed before starting
- Ask for the final report with artifacts, validation evidence and risks
What people use it for
- Planning and building a SaaS MVP across scope, implementation, testing and release steps
- Running a repeatable web application security review with explicit authorization and checkpoints
- Setting up an AI agent or LLM product with defined validation at each phase
- Stabilising browser end-to-end tests and flaky QA runs with a verified process
- Designing a domain model for a new service before implementation begins
Frequently asked questions
Is the Antigravity Workflows skill free?
The skill is listed under the MIT licence, which is stated in the directory entry. The reviewed files do not mention any price, so check the directory page for the current terms.
What do I need to install before using it?
You need the antigravity-workflows folder unzipped into ~/.claude/skills/ or a project's .claude/skills/, or the ZIP uploaded in the Claude apps. Most workflows also need companion skills such as ship-saas-mvp or qa-browser-automation present locally.
Why does the agent say a skill is missing?
The workflows depend on other skills being installed on your machine. The skill is designed to state the missing skill and fall back to the closest available one. Install the named companion skill, then run the request again.
Does the skill install other skills automatically?
No. Its rules say installation must be reviewed first with a dry run and only within your authorization. A failed preview never authorizes an install.
Does it run scripts or access the internet?
The listed requirements say it runs no scripts and uses no network access. Any work it coordinates through other skills may still depend on their own requirements, so review those too.
How does it compare with using a single skill directly?
The reviewed files do not compare it with other tools. It is most useful for multi-phase projects where you want a structured plan across several skills. For a one-off task, invoking the single relevant skill directly is simpler.
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 this skill if you have a multi-phase project, want a repeatable security review or browser QA process with explicit checkpoints, or would rather not pick every supporting skill yourself. Pick something else for simple questions or small edits, or if you expect it to work alone, since it depends on companion skills, credentials and infrastructure you must supply.
Download the skill