Skip to content

kbarbel640-del/skills / skills/0xnagato/antigravity-swarm

antigravity-swarm

Verified

Deploys autonomous sub-agents to perform tasks in the Antigravity IDE. Supports both manual dispatch and dynamic "Auto-Hiring" of agent teams.

FileKindSize
GEMINI_example.mdmarkdown455 B
README.mdmarkdown7.7 KB
README_KO.mdmarkdown9.2 KB
SKILL.mdskill md4.5 KB
_meta.jsonconfig288 B
gemini-extension.jsonconfig210 B
package.jsonmanifest360 B
requirements.txtmanifest25 B
scripts/compactor.pyscript2.9 KB
scripts/dispatch_agent.pyscript7.0 KB
scripts/orchestrator.pyscript10.3 KB
scripts/planner.pyscript10.1 KB
scripts/ultrawork_loop.pyscript2.3 KB
# Antigravity Sub-Agents Skill ๐Ÿš€

**Hire a team of AI agents to code for you.**

This skill allows you to spawn multiple specialized AI agents (Sub-Agents) to work on complex tasks in parallel. Whether you are using the terminal or the Antigravity IDE, this tool orchestrates a team of expertsโ€”Architects, Engineers, and Validatorsโ€”to solve your problems efficiently.

```text
โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”“
โ”ƒ                   โœจ Antigravity Swarm Mission Control โœจ                    โ”ƒ
โ”—โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”โ”›
โ•ญโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ฌโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ฌโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ฌโ”€โ”€โ”€โ”€โ”€โ”€โ•ฎโ•ญโ”€โ”€โ”€โ”€โ”€ Live Activity โ”€โ”€โ”€โ”€โ”€โ•ฎ
โ”‚ Agent            โ”‚ Role      โ”‚   Status    โ”‚ Time โ”‚โ”‚                         โ”‚
โ”œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ผโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ผโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ผโ”€โ”€โ”€โ”€โ”€โ”€โ”คโ”‚                         โ”‚
โ”‚ Prometheus       โ”‚ serial    โ”‚ โœ” Completed โ”‚ 32s  โ”‚โ”‚                         โ”‚
โ”‚ Junior           โ”‚ serial    โ”‚ ๐Ÿ”„ Running  โ”‚ 15s  โ”‚โ”‚ Agent: Junior           โ”‚
โ”‚ Quality_Validatโ€ฆ โ”‚ validator โ”‚ โณ Pending  โ”‚ -    โ”‚โ”‚ Action: Implementing    โ”‚
โ•ฐโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ดโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ดโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ดโ”€โ”€โ”€โ”€โ”€โ”€โ•ฏโ”‚ core logic...           โ”‚
                                                     โ”‚                         โ”‚
                                                     โ•ฐโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ•ฏ
```

> [!IMPORTANT] > **Why do you need this?**
> As of 2026-01-31, **there is NO native method** to deploy sub-agents in Gemini CLI or the Antigravity IDE.
> This skill bridges that gap, unlocking true parallel agent orchestration for your environment.

> [!NOTE] > **Windows Compatibility**
> This skill includes native support for Windows PowerShell environments (CP949/Korean locale) by enforcing UTF-8 encoding for all I/O operations.

> [!WARNING]
> **Do NOT modify files in this directory while the Orchestrator is running.**
> The system actively reads and writes to `task_plan.md`, `findings.md`, and `subagents.yaml`.
> Manual edits during execution may cause race conditions or inconsistent agent behavior.

> [!NOTE]
> **State Files Management**
> When you initiate a major mission, the system will automatically generate `task_plan.md`, `findings.md`, and `progress.md` in your project root.
> These files serve as the "shared memory" for the agent swarm. Do not delete them while a mission is active.

---

## Why use this?

- **Parallel Execution**: Why wait for one agent to finish? Run 3 agents at once (e.g., one writes docs, one writes tests, one writes code).
- **Specialization**: Assign different "Personas" (Prompts) to each agent.
- **Quality Assurance**: **[NEW]** Every team includes a mandatory **Validator Agent** who reviews the work before completion, ensuring high quality.

---

## Installation

1.  **Locate the Skill**: Ensure this repository is in your skills folder:
    ```bash
    ~/.gemini/skills/antigravity-swarm/
    ```
2.  **Install Dependencies**:

    ```bash
    pip install -r requirements.txt
    ```

    _(Requires `rich`, `pyyaml`, etc.)_

3.  **Gemini CLI**: Ensure the `gemini` command is installed and in your PATH.

---

## User Manual

There are three main ways to use this skill.

### ๐Ÿ…ฐ๏ธ Scenario A: Gemini CLI (Manual Mode)

Use this when you want to manually "hire" a team to do a job for you from your terminal.

**Step 1: Hire the Team (Plan)**
Run the planner with your mission description.

```bash
python3 scripts/planner.py "Create a Snake game with a green background and score tracking"
```

**Step 2: Review the Plan**
The Planner will propose a team from the **Available Agent Pool** (see below) and draft a `task_plan.md`.

**Step 3: Execute (Orchestrate)**
Once approved, launch the team:

```bash
python3 scripts/orchestrator.py
```

### ๐Ÿ…ฑ๏ธ Scenario B: Ultrawork Loop (Autonomous Mode)

Use this for "set and forget" operations. The system will Plan -> Act -> Verify -> Fix in a loop until the mission succeeds.

```bash
python3 scripts/ultrawork_loop.py "Refactor the authentication module and add tests"
```

### ๐Ÿ†Ž Scenario C: Antigravity IDE (Agent Integration)

If integrated into `~/.gemini/GEMINI.md`, the Main Agent will **automatically** trigger this skill for complex tasks.

---

## Available Agent Roles (Oh-My-Opencode Standard)

The Planner automatically selects the best experts for your mission from this pool:

| Role | Expertise | Model |
|------|-----------|-------|
| **Oracle** | Architecture, Deep Debugging, Root Cause Analysis | gemini-3-pro |
| **Librarian** | Documentation, Code Structure, Research | gemini-3-flash |
| **Explore** | Fast Search, Pattern Matching, Reconnaissance | gemini-3-flash |
| **Frontend** | UI/UX, Styling, Accessibility | gemini-3-flash |
| **Multimodal** | Vision Analysis, Mockups | gemini-3-pro |
| **Doc_Writer** | READMEs, API Docs, Comments | gemini-3-flash |
| **Prometheus** | Strategic Planning, Requirements Gathering | gemini-3-pro |
| **Momus** | Critical Review, Risk Identification | gemini-3-pro |
| **Sisyphus** | Task Coordination, Delegation (PM) | gemini-3-flash |
| **Junior** | Implementation, Coding, Bug Fixing | gemini-3-flash |
| **Quality_Validator** | Final Verification & Testing (Mandatory) | gemini-3-flash |

---

## Protocol & Architecture

### The "Manus Protocol"

This skill enforces a rigorous state management protocol:

- **`task_plan.md`**: The master checklist.
- **`findings.md`**: Shared scratchpad for agents to exchange knowledge.
- **`progress.md`**: Immutable log of what has been done.

### The "Validator" Rule

By design, the **final agent** in any generated team is strictly enforced to be a **Quality Validator**.

- **Role**: Reviewer / QA.
- **Task**: Check if `task_plan.md` is complete, run tests, and ensure the user's requirements are met.
- **Benefit**: Redundant self-correction loop.

---

## Directory Structure

```text
antigravity-swarm/
โ”œโ”€โ”€ scripts/
โ”‚   โ”œโ”€โ”€ planner.py       # The "Hiring Manager" (Generates subagents.yaml)
โ”‚   โ”œโ”€โ”€ orchestrator.py  # The "Project Manager" (Runs the agents)
โ”‚   โ”œโ”€โ”€ dispatch_agent.py # The "Worker" (Shim layer for Gemini CLI)
โ”‚   โ””โ”€โ”€ ultrawork_loop.py # The "Autonomous Loop" (Plan -> Act -> Verify)
โ”œโ”€โ”€ subagents.yaml       # The current team roster
โ”œโ”€โ”€ task_plan.md         # Current mission status
โ””โ”€โ”€ README.md            # This file
```

---

## โญ๏ธ Credits & Inspiration

This project is heavily inspired by **[Oh-My-Opencode](https://github.com/code-yeongyu/oh-my-opencode)**.

We have adopted its core philosophies:
- **Multi-Agent Orchestration**: Using specialized roles (Oracle, Librarian, Sisyphus) for distinct tasks.
- **The "Manus Protocol"**: Using persistent markdown files (`task_plan.md`, `findings.md`) for state management. This pattern is heavily inspired by **[planning-with-files](https://github.com/OthmanAdi/planning-with-files)**.
- **Ultrawork Loop**: The autonomous "Plan -> Act -> Verify" cycle.

Huge thanks to the original creators for defining these agent interaction patterns.
writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deploymentStars: 323kUpdated Claude CodeCodexVerifiedInstall
xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path โ€” even casually (like "the xlsx in my downloads") โ€” and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release.Stars: 118kUpdated Claude CodeCodexVerifiedInstall