Antigravity Skills
Four Google Antigravity slash commands: bug investigation, a duel of two solutions, speed measurement and context compaction.
- 61 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- `/battle` creates branches and merges the result, and `/benchmark` writes and runs harness code on your machine.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/antigravity-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- `/battle` creates branches and merges the result, and `/benchmark` writes and runs harness code on your machine.
This is third-party code. Review the repository files before installing.
What it does
The skill set for Google Antigravity (`agy`) turns internal agent scenarios into slash commands. `/owl` splits a task between an investigator and a coder: the first tests hypotheses and boundary conditions, the second writes and re-verifies the solution. `/battle` creates two Git branches where rivals implement different approaches, and a referee hunts for flaws and merges the winner. `/benchmark` builds a standalone measurement harness, runs it headlessly and compares p50 and p99 latency with memory use. `/compaction` moves what matters from a long conversation into artifact files and flushes surplus token history.
Who it is for. Antigravity users fixing intermittent bugs and weighing architecture in large codebases who want conclusions verified by measurement rather than trust in the agent.
Good fit when
- A crash reproduces only sometimes and a patch on line 42 would just hide the problem.
- You must choose between a lock-free ring buffer and a mutex queue and there is no single answer.
- A hot-path optimization has to be confirmed with p50 and p99 figures rather than the agent words.
- A session has run for hours and the agent forgets early requirements.
Not a fit when
- The edit is small and two rival agents would only burn tokens.
- Work runs in another agent: the skills target the Antigravity directory.
- The repository cannot be split into Git worktrees, while `/battle` needs separate branches.
Example request
/owl Find why worker threads crash on connection reset, audit socket lifecycles and falsify hypotheses before writing code.Limitations
The skills target Google Antigravity 2.0 and go into `~/.gemini/config/skills/`, or into the project `.agents/skills/` when installed with the `-LocalWorkspace` flag. `/battle` runs two copies of the work on separate branches and merges one of them, so save current changes before starting. `/benchmark` writes and runs the harness code itself, and its results depend on the machine. The repository holds only four skills and install scripts, and the example prompts lean toward C++.
How to disable. Delete the `owl`, `battle`, `benchmark` and `compaction` folders from `~/.gemini/config/skills/` or the project `.agents/skills/`.
FAQ
How do I install the skills in Antigravity?
Paste a request in the Antigravity chat to install the skills from the repository into `~/.gemini/config/skills/`: the agent clones the repository and enables the commands. Or run `install.sh` by hand.
How do I install into one project only?
On Windows run `.\install.ps1 -LocalWorkspace` from the project root. The skills land in `.agents/skills/` and teammates get them with the repository.
How does /owl differ from a normal bug fix?
Two specialists work separately: one investigates hypotheses and invariants, the other writes code. That avoids closing a crash with a patch that breaks timers and events.
What does /battle do to the repository?
It creates two Git worktrees where rivals build different solutions. A referee tests both for flaws, then the winner is merged into the main branch.
Does it work in Claude Code?
The README describes the skills for Antigravity only, so compatibility with other agents is not claimed.
Discussion
No questions or reviews yet.
Related
SkillsSuperpowers for Google Antigravity: a fork of the methodology with native agy tools and no translation layer.
Installer of a working setup for Google Antigravity that puts rules, skills and workflows into a project with one npx command
19 skills for Claude Code, Codex and Antigravity CLI: Grok and Exa search, Context7 docs, Serena and task hand-offs between agents
An autonomous iteration skill for agents: modify, verify, revert failed changes, and repeat the loop toward a target metric
An Antigravity CLI plugin for engineering work: a separate design turn, documentation agents and edits only within a sealed plan
A skill that finds AI slop in code: duplicate helpers, empty abstractions, weak tests. It cleans one pattern per commit.
README in short
Installation takes one request to the agent or a PowerShell or Bash script. After installing, the commands are visible immediately, and a check is described in the Verification section.