@a5c-ai/babysitter-antigravity-cli

@a5c-ai/babysitter-antigravity

Babysitter integration package for Antigravity CLI.

Antigravity CLI is Google's successor to Gemini CLI. It uses a workflow-driven hook model and a SKILL.md plugin system. It is model-agnostic, supporting Gemini, Claude, and GPT providers.

This package ships an Antigravity CLI plugin bundle:

• SKILL.md — Antigravity-native skill manifest with YAML frontmatter
• plugin.json — Babysitter plugin manifest
• workflow.json — Workflow orchestration descriptor
• mcp.json — MCP config for multi-provider tool integration
• commands/ — Slash command definitions for all babysitter workflows
• hooks/ — Workflow-driven hook scripts
• bin/cli.js — babysitter-antigravity installer CLI

It uses the Babysitter SDK CLI and the shared ~/.a5c process-library state. The plugin registers workflow-driven hooks and commands so Antigravity CLI can drive the Babysitter orchestration loop from within the agent session.

Installation

Install the Babysitter CLI once:

npm install -g @a5c-ai/babysitter

Install the Antigravity plugin through the SDK helper:

# Global install
babysitter harness:install-plugin antigravity-cli

# Workspace install
babysitter harness:install-plugin antigravity-cli --workspace /path/to/project

You can also run the published package installer directly:

npx --yes @a5c-ai/babysitter-antigravity install --global
npx --yes @a5c-ai/babysitter-antigravity install --workspace /path/to/project

For development, use a symlink instead of copying files:

babysitter-antigravity install --symlink

After installation, restart Antigravity CLI to activate the plugin.

How It Works

The plugin implements a workflow-driven orchestration loop:

1. SessionStart fires when a new Antigravity CLI session begins. It ensures the correct SDK CLI version is installed (pinned via versions.json) and initializes session state under ~/.a5c/state/.

2. The SKILL.md context file is loaded into every session, instructing the agent on the full 8-step orchestration workflow -- from interviewing the user and creating a process definition through iterating effects and posting results.

3. The agent performs one orchestration phase per turn, then stops.

4. AfterAgent fires after every agent turn. It checks whether a babysitter run is bound to the current session. If the run is not yet complete, the hook returns {"decision":"block","reason":"...","systemMessage":"..."} to keep the session alive and inject the next iteration prompt. Once the agent emits <promise>COMPLETION_PROOF</promise>, the hook allows the session to exit.

Hook Types

Hook	Event	Type	Purpose
Session initialization	SessionStart	workflow	Installs the correct SDK version, creates session state
Continuation loop	AfterAgent	workflow	Blocks session exit and drives the orchestration loop until the run completes

Both hooks delegate to the SDK CLI via babysitter hook:run for all business logic. The workflow-driven hook model replaces the shell-hook scripts used by Gemini CLI.

Differences from Gemini CLI

Feature	Gemini CLI	Antigravity CLI
Hook model	Shell-hook scripts	Workflow-driven hooks
Plugin manifest	gemini-extension.json	SKILL.md (YAML frontmatter)
Model support	Gemini only	Gemini, Claude, GPT (model-agnostic)
Tool integration	Native	MCP-based multi-provider
Orchestration	Extension-based	Workflow + SKILL.md

Available Commands

All commands follow the orchestration workflow described in SKILL.md. Invoke them in Antigravity CLI with /babysitter:<command>.

Primary Orchestration Commands

Command	Description
/babysitter:call [instructions]	Start a babysitter-orchestrated run
/babysitter:plan [instructions]	Generate a detailed execution plan without running anything
/babysitter:yolo [instructions]	Start a run in fully autonomous mode
/babysitter:forever [instructions]	Start a run that loops indefinitely
/babysitter:resume [run-id]	Resume a paused or interrupted run

Diagnostic and Analysis Commands

Command	Description
/babysitter:doctor [run-id]	Run a health check on a run
/babysitter:retrospect [run-id...]	Analyze completed runs and suggest improvements

Lifecycle Management Commands

Command	Description
/babysitter:assimilate [target]	Convert an external methodology into native babysitter process definitions
/babysitter:cleanup [--dry-run]	Aggregate insights then remove old run data
/babysitter:observe	Launch the real-time observer dashboard

Setup Commands

Command	Description
/babysitter:user-install	First-time onboarding and user profile setup
/babysitter:project-install	Onboard a project for babysitter orchestration

Plugin and Community Commands

Command	Description
/babysitter:blueprints [action]	Manage Babysitter blueprints
/babysitter:contrib [feedback]	Submit feedback or contribute to the babysitter project
/babysitter:help [topic]	Show help for babysitter commands and workflows

Configuration

Source	Purpose
versions.json	Pins the required @a5c-ai/babysitter-sdk version
SKILL.md	Antigravity-native skill manifest with YAML frontmatter metadata
plugin.json	Babysitter plugin manifest declaring hooks, commands, and harness
workflow.json	Workflow orchestration descriptor for hook execution graph
mcp.json	MCP config for multi-provider tool integration
ANTIGRAVITY_SKILL_PATH env var	Path to the installed plugin root
BABYSITTER_LOG_DIR env var	Override the log directory (defaults to ~/.a5c/logs)
~/.a5c/state/	Session state directory
~/.a5c/user-profile.json	User profile for personalizing orchestration

Verification

Verify the installed plugin bundle:

babysitter-antigravity status --global

Verify the SDK CLI is available:

babysitter --version

Verify the active process-library binding:

babysitter process-library:active --json

Troubleshooting

SDK CLI not found after session start

The SessionStart hook installs the SDK automatically. If permissions prevent a global install, check the session-start log:

cat ~/.a5c/logs/babysitter-session-start-hook.log

If the CLI is still missing, install it manually:

npm install -g @a5c-ai/babysitter-sdk

Hook logs location

Log file	Contents
~/.a5c/logs/babysitter-session-start-hook.log	SessionStart hook output
~/.a5c/logs/babysitter-session-start-hook-stderr.log	SessionStart SDK stderr
~/.a5c/logs/babysitter-after-agent-hook.log	AfterAgent hook output
~/.a5c/logs/babysitter-after-agent-hook-stderr.log	AfterAgent SDK stderr

License

MIT