adk-agent

What problem does it solve?

This Skill prevents fragile Google ADK agent implementations by turning agent requirements into executable, testable documentation that must pass quality gates before deployment.

Core Features & Use Cases

• ADK-first project structure rules: Enforces a consistent layout (app/, tests/, deployment/, Dockerfile, pyproject.toml) so wiring and business logic stay clean.
• Typed agent definition patterns: Requires the root Agent + App + instruction prompt and tool schemas be fully specified, including typed tool signatures and docstrings.
• Agent TDD via evalsets: Converts spec-defined test cases into eval sets and an acceptance rubric, ensuring behaviors are validated by LLM-as-judge with defined thresholds.
• Serving, telemetry, and deployment constraints: Guides FastAPI serving, session types, privacy-safe telemetry, and Agent Engine deployment steps including requirements generation.
• Multi-agent routing patterns: Defines sub-agent orchestration plus a routing table for complex domains to reduce tool overload and improve reliability.

Quick Start

Create your ADK agent spec that includes the full instruction prompt, typed tool signatures with docstrings, and at least five eval cases mapped to TC IDs.