Claude Skill MartBrowse skillsQuick linesLearn by videoTerminal guideWhat is a Skill?
 
← Back to list

agy-worker (Antigravity delegation worker)

A workflow skill that delegates repo exploration and bounded implementation to Google Antigravity CLI (agy) while the driver keeps approval, diff review, and verification.

Dev & CodingAdvanced★ 416⑂ 43AI score 7/10Last updated: Sep 1, 2026

What it does

  • Routes requests into three workflows: explore (understand/review/plan), task (feature, refactor, tests, bounded repair), and project (broad build or audit-and-fix, up to 5 cycles).
  • Requires explicit, hash-bound approval of what leaves your machine: either whole-worktree approval (--approve-whole-worktree + manifest SHA256) or scoped transmission (--provider-scope + --approve-transmission-sha).
  • Enforces a verification loop: read the real Git diff, pick your own build/test/lint commands (never run commands_run from the envelope), run writable checks in an isolated copy, then finalize as verified, partially_verified, rejected, or blocked.
  • Defines hard stops for unapproved transmission, secrets or denied paths inside staged content, writes escaping the worktree or touching .git, permission-bypass flags, and unauthorized commits/pushes/releases.

Who it's for

  • Developers driving OpenAI Codex CLI who want to hand off heavy repo work to Google Antigravity (agy).
  • Teams that must document data boundaries and approval trails when source code is sent to an external agent.
  • Note: the skill states Claude and Claude Code hosts are not supported; it needs Bash, Python 3, git, agy, and provider network access.

Usage examples

  1. Onboarding to an unfamiliar repo: run explore for a structural map, then spot-check the material claims yourself.
  2. Bounded change: run task to add tests or refactor a module, inspect the diff, and validate with your own test command in a verification copy.
  3. Project-wide audit: run project, review repo-wide changes, run build/test/lint, and have a separate agent or fresh context accept the final result.

· · · Install guide · · ·

Try it now, no install

Paste this into Claude to use the skill without installing anything.

Read the instructions in this file and follow them to help me:
https://raw.githubusercontent.com/aiskillstore/marketplace/HEAD/pending/cagdasyurekli/agy-worker/SKILL.md

What I want: (describe your task here)

If Claude can't open the link, open it yourself and paste the contents instead.

↓ If it works for you, download the ZIP below and install it. Then it runs on its own — no pasting each time.

Install in the Claude app (no terminal)
  1. Download the ZIP with the button below.
  2. In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
  3. Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.
↓ Download ZIP
Install in Claude Code

Let Claude do it — paste this into Claude Code

Install the skill I found on Claude Skill Mart.
Copy the pending/cagdasyurekli/agy-worker folder from the GitHub repo aiskillstore/marketplace into my ~/.claude/skills/agy-worker/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/aiskillstore/marketplace.git && mkdir -p ~/.claude/skills && cp -r marketplace/pending/cagdasyurekli/agy-worker ~/.claude/skills/

⚠ This is a third-party skill. Check the source repository before installing.

  1. Open a terminal in your working directory.
  2. Clone the repo: git clone https://github.com/aiskillstore/marketplace.git
  3. Copy the skill: mkdir -p ~/.claude/skills && cp -r marketplace/pending/cagdasyurekli/agy-worker ~/.claude/skills/
  4. Verify prerequisites: bash --version, python3 --version, git --version, and that agy --help works.
  5. Because the skill targets OpenAI Codex CLI and explicitly does not support Claude Code, install it into your Codex CLI skill/prompt path for real use.
  6. Run the offline check: resolve the package path via scripts/resolve-pipeline.sh, then run "$PIPELINE/doctor.sh" --repo /absolute/path. A ready result covers offline prerequisites only.
  7. On first use, always start with run --preview to inspect what would be transmitted, then re-run with the exact approval digest flag.
View source on GitHub ↗