Antigravity CLI Replaces Gemini — AI Dev Pulse · Jun 18, 2026

Google’s consumer-facing Gemini CLI reaches end-of-life today, June 18, forcing developers on Pro, Ultra, and free tiers to migrate workflows to the new Antigravity CLI. The change consolidates Google’s agent tooling into a Go-based platform with shared harness capabilities across desktop and terminal surfaces. Meanwhile, Snap’s new SPECS AR glasses arrive with production-grade developer tooling that lets teams build spatial Lenses directly inside the same AI coding environments they already use daily. Security researchers are also highlighting a sharp rise in “agentjacking” exploits that abuse common error-handling patterns in Claude Code, Cursor, and Codex. These updates land against a backdrop of continued policy friction around frontier models and novel deployment stories such as Gemma 3 operating in orbit. For builders, the practical questions center on migration friction, new spatial AI surfaces, and hardening agentic code against prompt-based attacks.

Top Stories

Google sunsets Gemini CLI for consumer users; Antigravity CLI goes live Practical dev impact: Teams relying on Gemini CLI for terminal agent orchestration must migrate scripts, hooks, and skills to the new Antigravity CLI command set before today’s cutoff or face broken CI/CD and local workflows.

Snap SPECS AR glasses ship with Claude Code, Codex, and Cursor integrations Practical dev impact: Lens Studio now supports agentic development workflows inside the three most-used AI coding tools, letting developers prototype, test, and publish spatial experiences without switching IDEs.

Agentjacking attacks exploit Claude Code, Cursor, and Codex at scale Practical dev impact: Developers should add input sanitization and Sentry-like error validation checks to any agent that processes external error messages, as fake error payloads now achieve 85% success rates against popular coding agents.

Gemma 3 runs as first LLM in orbit Practical dev impact: Edge and constrained-environment deployments gain a new reference point; teams working on satellite or remote hardware can benchmark against a production-grade open model already proven in space.

Practical Impact Analysis

Recommended Tutorial Idea

#!/bin/bash
# migrate-gemini-to-antigravity.sh
# Usage: ./migrate-gemini-to-antigravity.sh "your task description"

set -euo pipefail

TASK="$1"

# Old Gemini CLI pattern (will fail after cutoff)
# gemini agent run --task "$TASK"

# New Antigravity CLI with validation wrapper
if [[ "$TASK" =~ (error|exception|traceback) ]]; then
  echo "Potential agentjacking payload detected. Sanitizing..."
  TASK=$(echo "$TASK" | sed 's/[^a-zA-Z0-9 .,!?-]//g')

... click "Show full code" below to expand

#!/bin/bash
# migrate-gemini-to-antigravity.sh
# Usage: ./migrate-gemini-to-antigravity.sh "your task description"

set -euo pipefail

TASK="$1"

# Old Gemini CLI pattern (will fail after cutoff)
# gemini agent run --task "$TASK"

# New Antigravity CLI with validation wrapper
if [[ "$TASK" =~ (error|exception|traceback) ]]; then
  echo "Potential agentjacking payload detected. Sanitizing..."
  TASK=$(echo "$TASK" | sed 's/[^a-zA-Z0-9 .,!?-]//g')
fi

agy run --task "$TASK" --plugin security-validator --async

Replace legacy `gemini` invocations with `agy`, add the simple regex guard shown above, and test the script against both benign and malicious error strings before production rollout.

Grok Deep Dive