Skip to content
SkillScript-led5,255 tokens in use

Gemini Agents

Drive Google's Antigravity CLI as a requested peer, builder or second reviewer, only after the user has agreed and heard the support status.

680 stars on the repository MIT Licensephuryn/product-engineering · agent-teams/skills/gemini-agents
npx skills add phuryn/product-engineering
View source

Install command as written in the repository's README

Skill label

measured from its files

What installing Gemini Agents costs

44tokens idle
5,255tokens in use
7,856tokens, up to

Idle is what sits in context while the skill is installed. In use is its instructions, loaded when it runs. Up to adds the files its instructions link. Counts are approximate, made offline with cl100k_base.

Model dependence Script-led

Script-ledMixedModel-led

Most of this skill is scripts, so it behaves much the same whichever model runs it.

This describes the skill's structure, how much of it is scripts. It does not say how well it works. Scripts: scripts/agy.py, scripts/test_agy.py.

Touches

shelldetectednetworkdetectedwrites filesdetectedgitdetectedcredentialsdetected

Declared tags come from the skill's frontmatter; detected ones from its scripts, its code blocks and the commands it names in inline code. This is detected, not audited.

How it scores

Quality

How well the skill is written, counted from its own files: its frontmatter, its steps, examples, stated output and constraints, its size and whether every file it ships is one it names.How we score it →
Good
  • Frontmatter8/15
  • Structure19/25
  • Examples12/15
  • Output15/15
  • Size13/15
  • Resources10/15

Security

Starts at 100 and loses points for each thing in the skill's files that a reader should look at first.How we score it →
Good
  • Runs shell commands−5

    does this on the machine, and the frontmatter does not say so

    SKILL.md, line 72
  • Writes files−8

    does this on the machine, and the frontmatter does not say so

    SKILL.md, line 72
  • Reads credentials−15

    does this on the machine, and the frontmatter does not say so

    scripts/agy.py, line 200
  • Reaches the network−10

    does this on the machine, and the frontmatter does not say so

    scripts/agy.py, line 1450

Repository health

Pre-computed score based on development activity, maintenance, community, maturity, and trend momentum.How we score it →
Good
  • Development activity100
  • Maintenance100
  • Community44
  • Maturity4
  • Momentum40

Gemini Agents runs Google’s Antigravity CLI (agy) as a partner when you ask for it: “ask Gemini” or “have Gemini build or review this”. It is never a default. Google’s FAQ, read by the author on 2026-11-13, says driving agy from another agent on a consumer plan is unsupported and can risk suspension, so the skill requires the user’s clear agreement before the first run.

A Python wrapper, agy.py, sets the boundary for each turn, checks for repository hooks and plugins, probes whether its denials hold, and judges the events a turn produced.

What You Get

  • agy.py, a Python wrapper with consult, build and review modes plus —check, —probe, —dry-run and —replay
  • A user-agreement step with the support status stated before any run
  • A refusal to start when workspace hooks, MCP configs or plugins are present unless the owner allows it
  • Recorded events and metadata so a run can be replayed and judged
  • Dated notes on what the sandbox allowed and refused on the tested platform, and a list of what is not established

Common Use Cases

  • Getting a requested second opinion from another model family
  • Letting Gemini build a stage in its own clone with the sandbox on
  • Reviewing a revision in a detached clone as an extra check beside the release reviewer
  • Probing whether the CLI’s permission boundaries still hold after an update

What Makes It Unique

Most partner skills assume the CLI behaves. This one records what the CLI did when tested on one day, including a denied permission that ended a turn with exit 0 and an empty answer, and encodes those findings as checks in the wrapper. It starts with a consent gate, and no other skill in the collection names Gemini as a default. Its Not established list names untested routes, such as Linux, Windows and API-key sign-in.

Under The Hood

The skill is instructions plus a Python wrapper and its test file. When you agree to a run, the wrapper launches the agy command-line tool as a subprocess in your repository or a detached clone, passes the brief as a file, and keeps session state under the Gemini folder in your home directory. In build mode Gemini can write files and run shell commands inside its sandbox. Consult mode can still use web search, and review keeps shell writes inside the clone. The wrapper runs git commands to list files, clone and compare status, and refuses to start when it finds workspace hook or plugin entries. The check command looks at the environment and the CLI’s settings to report which sign-in route is in use, without printing credentials, and the probe command tries a network request from the sandboxed shell to confirm it is denied. The owner puts any API key in the environment; the agent never asks for one.

Join founders building with open source

Opinionated takes, migration guides, cost-saving tips, and insights from the open source ecosystem.

Subscribe on SubstackJoin 750+ subscribers