From adk-agent-builder
Use this skill to scaffold an ADK 2.0 Workflow Agent — Sequential, Loop, or Parallel — for deterministic multi-step pipelines. Triggers on: "ADK sequential agent", "build an ADK pipeline", "loop agent that retries", "run agents in parallel", "ADK workflow agent", "chain ADK agents", "deterministic agent flow". Generates SequentialAgent / LoopAgent / ParallelAgent code wiring N sub-LlmAgents in the requested topology, with termination conditions for loops. Use when control flow is fixed and deterministic — for code-driven branching/iteration use dynamic-workflow-builder, for graph-routed workflows use graph-workflow-designer.
How this skill is triggered — by the user, by Claude, or both
Slash command
/adk-agent-builder:workflow-agent-scaffoldThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Scaffold deterministic workflow agents in ADK 2.0.
Scaffold deterministic workflow agents in ADK 2.0.
| Class | Use when |
|---|---|
SequentialAgent | Steps must run in order; output of step N feeds step N+1 |
LoopAgent | Repeat a sub-agent until a termination condition fires |
ParallelAgent | Run independent sub-agents concurrently and collect results |
from google.adk.agents import LlmAgent, SequentialAgent
researcher = LlmAgent(
name="researcher",
model="gemini-2.5-flash",
instruction="Research the topic and write findings to state['research'].",
output_key="research",
)
writer = LlmAgent(
name="writer",
model="gemini-2.5-pro",
instruction="Using state['research'], write a 500-word brief.",
output_key="brief",
)
root_agent = SequentialAgent(
name="research_then_write",
sub_agents=[researcher, writer],
)
from google.adk.agents import LlmAgent, LoopAgent
drafter = LlmAgent(
name="drafter",
model="gemini-2.5-flash",
instruction="Draft a response. If state['quality'] != 'approved', revise.",
output_key="draft",
)
root_agent = LoopAgent(
name="iterative_drafter",
sub_agents=[drafter],
max_iterations=5,
)
from google.adk.agents import LlmAgent, ParallelAgent
news = LlmAgent(name="news_search", model="gemini-2.5-flash", output_key="news")
weather = LlmAgent(name="weather_check", model="gemini-2.5-flash", output_key="weather")
stocks = LlmAgent(name="stock_check", model="gemini-2.5-flash", output_key="stocks")
root_agent = ParallelAgent(
name="morning_briefing",
sub_agents=[news, weather, stocks],
)
nameoutput_keys don't collidemax_iterations is set (avoid infinite loops)Guides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.
npx claudepluginhub healthcare-ai-consulting-llc/adk-2-toolkit --plugin adk-agent-builder