Skills DirectorySkills Directory
SkillsLearnSecurityCategoriesDocsCommunityBlog
Sign InSubmit Skill
Skills Directory

Security-tested agent skills for Claude, coding agents, and AI workflows.

Directory

  • Browse Skills
  • All Skills A–Z
  • Claude Skills
  • Claude Code Skills
  • Agent Skills
  • Categories
  • Authors
  • Submit a Skill

Learn

  • Learn Hub
  • Install Claude Skills
  • Write SKILL.md
  • Skills vs MCP
  • Directories Compared

Security

  • Security
  • Methodology
  • Secure Claude Skills
  • Security Badges

Company

  • About
  • Community
  • Blog
  • API Docs
  • Advertise

2026 Skills Directory. All rights reserved.

ProTermsPrivacyRefunds
Back to skills

Agy

ASecurity

Delegate coding work to the Antigravity `agy` CLI agent (Google Antigravity, non-interactive print mode). Use this whenever the user says "agy", "antigravity", "antigravity se karwa", "spawn antigravity", "delegate to antigravity", asks you to "make antigravity build/fix/scaffold X", wants to change the antigravity model, or wants a large build/scaffold/refactor task offloaded to another coding agent while you orchestrate. Also use it to generate raster images — icons, illustrations, hero ima...

2 stars
0 votes
0 copies
0 views
Added 9/28/2026
ai-agentsgoshellbashsqlgitdatabase

Works with

climcp

Security Analysis

A100/100

Scanned 9/28/2026

Install to Claude Code

$npx -y skills add shauryaaojha/claude-code-delegate --skill agy --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Agy?

Add the live security badge to your README — it updates automatically with every re-scan.

Security grade badge for Agy
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/shauryaaojha-agy/badge)](https://www.skillsdirectory.com/skills/shauryaaojha-agy)

More formats (shields.io, HTML) on the badges page.

Files
SKILL.md
---
name: agy
description: Delegate coding work to the Antigravity `agy` CLI agent (Google Antigravity, non-interactive print mode). Use this whenever the user says "agy", "antigravity", "antigravity se karwa", "spawn antigravity", "delegate to antigravity", asks you to "make antigravity build/fix/scaffold X", wants to change the antigravity model, or wants a large build/scaffold/refactor task offloaded to another coding agent while you orchestrate. Also use it to generate raster images — icons, illustrations, hero images, placeholder art, "image bana do", "generate an image/logo/icon" — since agy has a built-in image generation tool and you do not. Also use it when the user has previously said coding should go through antigravity and a new implementation task comes up.
---

# agy — delegating coding tasks to Antigravity

`agy` is the Antigravity CLI. In print mode (`-p`) it runs one autonomous coding turn in
a workspace and exits. You stay the orchestrator: you write the spec, spawn agy, verify
its output (build/tests/diff), and fix or re-prompt. agy does the bulk of the typing.

## The command

```bash
agy -p "$(cat "<task-file>")" \
  --model gemini-3.8-flash-high \
  --dangerously-skip-permissions \
  --print-timeout 50m \
  --add-dir "<absolute-project-dir>"
```

Or use the bundled runner, which does the same thing with less quoting risk:

```bash
bash ~/.claude/skills/agy/scripts/run_agy.sh <task-file> <project-dir> [model] [timeout]
```

Why each flag matters:
- `-p` / `--print` — non-interactive; without it agy opens a TUI and hangs your shell.
- `$(cat task-file)` — put the prompt in a file, never inline. Multi-line prompts with
  backticks, `$`, quotes or heredocs break bash quoting (`unexpected EOF`) and PowerShell
  mangles them worse. Write the file with the Write tool, then `cat` it.
- `--model` — pick explicitly (see below). Ask the user only if they haven't stated a
  preference in this session; otherwise reuse the last one they named.
- `--dangerously-skip-permissions` — agy auto-approves its own tool calls. Without it the
  run blocks on prompts nobody can answer. Confirm with the user the first time in a
  session; after that it's implied.
- `--print-timeout 50m` — default `0` waits forever; a stuck run would block you. 50m covers
  a full scaffold. Use the Bash `timeout` param (max 600000 ms) with `run_in_background: true`
  for anything over a few minutes so you aren't blocked.
- `--add-dir` — agy's workspace root. It **creates files relative to this dir**, so point it at
  the exact folder you want output in (see "Where files land").
- Do NOT pass `--effort` with Claude models — `--effort is not supported for model "claude-opus-4-6-thinking"`.
  Gemini models encode effort in the model id (`-high/-medium/-low`) instead.

## Models

`agy models` lists what's available. Known ids (Sept 2026):

| id | use for |
|---|---|
| `gemini-3.8-flash-high` | default for builds/scaffolds — fast, generous quota |
| `gemini-3.1-pro-high` | harder reasoning, slower |
| `claude-opus-4-6-thinking`, `claude-sonnet-4-6` | strongest, but quota is small and runs out mid-task; no `--effort` |
| `gpt-oss-120b-medium` | cheap small edits |

If a run dies with a quota/rate error, switch model rather than retrying the same one. If the
user says "model change karde X pe", use X for every subsequent run in the session.

## Writing the task file

agy starts cold — it has none of your conversation context. The task file is its entire
brief, so write it like a handoff to a competent contractor who has never seen the repo:

1. **Where** — absolute working dir on the first line (`You are working in C:/…/project`).
2. **Read first** — exact paths of spec, existing code and conventions it must match.
3. **Deliver** — numbered list of files/features with concrete acceptance criteria
   (`npm run build` and `npm test` must pass; function signatures stay identical; etc.).
4. **Don't** — what is *not* its job (e.g. "do not connect to any database, the orchestrator
   applies SQL via MCP"; "do not install X"). agy will otherwise wander into it.
5. **Finish** — "End with a concise summary: files written, build/test results, anything
   you were unsure about." That summary is what comes back on stdout; it's your only view
   into what happened besides the diff.

Keep decisions *made*, not open — agy can't ask you questions in print mode, it will guess.
Put the file in the project (e.g. `<area>/AGY_TASK.md`) if the user may want to see/reuse it,
otherwise in the scratchpad.

See `references/task-template.md` for a fill-in template.

## Where files land (Windows gotchas)

- agy scaffolds *inside* `--add-dir`. If you want `project/app/` and pass `--add-dir project`,
  you may get `project/app/` or agy may pick a name; check with `ls` and `mv` after.
- Moving the output folder can fail with `Device or resource busy` when your shell's cwd is
  inside it (or a dev server/agy is still running). `cd` out (or run `mv` with absolute paths
  from another dir) and retry.
- Paths in the task file: use forward slashes (`C:/Users/…`) — both agy and Git Bash accept them.

## After the run — always verify

agy reports success optimistically. Before telling the user it's done:

```bash
git status --short          # what it actually touched
npm run build && npm test   # or the project's equivalent
```

Typical leftovers to fix yourself (faster than a re-prompt): unused imports (`TS6133`),
stubbed helpers left as TODO, `.env` committed, wrong folder. Re-prompt agy (new task file,
same command, or `agy -c -p "…"` to continue the last conversation) only for substantial gaps.

If agy's quota dies mid-task, finish the remaining items yourself — don't leave the user with
a half-built tree.

## Generating images

agy has a built-in `generate_image` tool; you don't. When a task needs a real raster image
(app icon, illustration, hero image, placeholder art), send just that image to agy
instead of drawing it with SVG or canvas code.

```bash
bash ~/.claude/skills/agy/scripts/gen_image_agy.sh <prompt-file> <out-dir>/<name>.png [model] [timeout]
```

- **The prompt file describes only the picture:** subject, style, composition, colours,
  what to leave empty, and "no text" unless you want lettering. The script adds the
  instructions agy needs: use `generate_image`, don't draw it with code, save under
  exactly this name, touch nothing else.
- **It runs agy inside the output folder**, because agy saves relative to its workspace.
  It refuses to overwrite an existing file.
- **It checks the file, not the summary.** It exits `2` if nothing was saved and `3` if
  the file isn't a PNG/JPEG/WebP (usually agy fell back to writing an SVG). On a
  non-zero exit, re-run once. Don't paper over it.
- **Look at the result before using it.** Read the PNG yourself. Image models misspell
  lettering and ignore negative instructions, so check the details the task depends on.
- Output is 1024×1024 PNG by default, around 800 KB. Resize or compress for the web
  yourself; don't ship it raw as a favicon.
- One image per run. For a set (e.g. five feature illustrations), launch one run per image
  with `run_in_background: true` rather than asking one run for all five, so one failure
  doesn't cost the whole set.
- Tested with `gemini-3.8-flash-high`. Keep the default model unless it refuses or fails.

Inside a larger build task, you can also just write "generate `public/hero.png` with your
image generation tool — do not draw it with code" in the task file. Afterwards, check it's
a real image the same way (`file public/hero.png`).

## Interactive / other modes

- `agy -i "<prompt>"` starts interactive with a first prompt — only if the user wants to drive
  agy themselves; tell them to run it via `! agy -i …`.
- `agy -c -p "…"` continues the most recent agy conversation (keeps its context).
- `agy --mode plan -p "…"` for a plan-only pass when the task is fuzzy.
- `--output-format json` if you need to parse the result programmatically.

Attribution

shauryaaojhashauryaaojha
View sourceMore from shauryaaojha →
SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments (0)

No comments yet. Be the first to comment!

SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Related Skills

Caveman

Ultra-compressed communication mode that cuts output tokens while keeping technical accuracy. Levels: lite, full, ultra and the wenyan variants. Use for /caveman, "caveman mode", "talk like caveman", "be brief" or "less tokens".

1074701 votes

Hyperplan

Adversarial multi-agent planning skill. Self-orchestrates 5 hostile category members (unspecified-low, unspecified-high, deep, ultrabrain, artistry) via team-mode for ruthless cross-critique debate, distills only the defensible insights, then MANDATORILY hands the distilled insight bundle to the `plan` agent for executable plan formalization. Use when planning needs maximum rigor and surfacing of weak assumptions, blind spots, and over-engineering. Triggers: 'hyperplan', 'hpp', '/hyperplan', ...

695601 votes

Mcp Code Execution

Routes multi-tool workflows through MCP servers for large datasets and pipelines. Use when Bash tool overhead is limiting throughput on data-heavy tasks.

3351 votes

catchup

Recovers the conversation and failed tool calls of a previous Codex, Claude Code, Antigravity, Cline, Copilot CLI, Cursor, DeepSeek Harness, Kimi, OpenCode, Pi Agent, or ZCode session. Use when the user says "catch up", "what did the last session do", "get me up to speed", "I switched agents", asks to recover/summarize a previous session before continuing, or asks to diagnose or report a catchup failure. Do NOT use for the current conversation, git history, or any non-agent log.

691 votes

math-skill

A comprehensive mathematical reasoning skill for AI assistants — handles arithmetic to research-level problems with rigorous step-by-step reasoning, systematic verification, and transparent uncertainty handling

381 votes
View all in ai-agents →