Android CLI and Skills for AI Agent Development Workflows

Google introduced Android CLI and Android Skills, tools designed to improve AI agent workflows for Android development outside of Android Studio.
Android CLI Features
The revitalized Android CLI serves as a lightweight, programmatic interface for terminal-based Android development. Key commands include:
android sdk install- downloads specific SDK components for lean environmentsandroid create- generates new projects from official templates with recommended architectureandroid emulator- creates and manages virtual devicesandroid run- deploys apps to devicesandroid update- ensures latest capabilities are available
Internal experiments showed Android CLI reduced LLM token usage by more than 70% and completed tasks 3x faster compared to using standard toolsets alone.
Android Skills Repository
The Android Skills GitHub repository provides modular, markdown-based instruction sets (SKILL.md files) that trigger automatically when prompts match skill metadata. Initial skills include:
- Navigation 3 setup and migration
- Implementing edge-to-edge support
- AGP 9 and XML-to-Compose migrations
- R8 config analysis
Skills can be browsed and set up using the android skills command when using Android CLI.
📖 Read the full source: HN AI Agents
👀 See Also

LobsterBoard adds theme system and template gallery
LobsterBoard now includes a theme system with five visual options and a template gallery that allows users to export and import dashboard layouts with automatic sensitive data stripping.

Why Deterministic Workflows Outperform AI-Driven Orchestration for Agent Systems
A developer with a year of experience building agent systems shares that AI-driven orchestration failed reliably due to non-deterministic routing, compounding errors, cost explosion, and impossible debugging. Switching to deterministic workflows with code-based orchestration eliminated orchestration failures.

Automating Datadog Alert Triage with Claude Code and MCP
A developer built a system using Claude Code skills and Datadog's MCP server to automatically check monitoring alerts, classify issues, and open fix PRs via cron job. The setup takes about 30 minutes and runs parallel AI agents in isolated worktrees.

Pilot Shell: A Structured Workflow Layer for Claude Code
Pilot Shell adds spec-driven TDD workflows, quality hooks, context engineering, and token optimization on top of Claude Code — without the complexity of multi-agent frameworks.