Level 1: Certified ArcFlow Workflow Developer

Title: Certified ArcFlow Workflow Developer

Prerequisite: None. Start here if you write agent workflows in Python or TypeScript.

What certified means at this level

You can build and run linear multi-step workflows with real or stub providers, configure agents with instructions and tools, read execution traces for basic lifecycle events, use arcflow init and arcflow run from the CLI, and diagnose WorkflowConfigurationError and WorkflowExecutionError without external help.

Competencies covered

Competency	Demonstration
Linear workflows	Three or more ordered steps with handoff
Agent configuration	Instructions, role, optional tools
Providers	Stub for tests; switch to one real LLM provider
Traces	Identify `WorkflowStarted`, `StepCompleted`, `WorkflowCompleted`
CLI	`init` project and `run` a script locally
Errors	Fix config errors vs execution failures

Continue with

Topic	Document
Mental model	What is ArcFlow
First run	First workflow in five minutes
SDK setup	Python quickstart
Workflows	Linear workflows
Agents	Defining agents
Providers	Provider configuration
Traces	Execution traces
Tutorial	Track A

Practical project

Build a three-step research, analyze, summarize pipeline:

Step	Agent role	Output expectation
1 Research	Gather facts on a topic you choose	Structured notes
2 Analyze	Identify themes and risks	Analysis text
3 Summarize	Executive summary	Final string returned from `run()`

Requirements

Requirement	Detail
Real LLM provider	At least one step uses a non-stub provider via env configuration
Tool	At least one agent registers a simple tool (calculator, datetime, or domain stub)
Trace verification	Script asserts required lifecycle event kinds after `run()`
CLI	Project creatable via `arcflow init`; workflow runnable via CLI
README	Lists env vars and verify command

Suggested starting points

Resource	Link
Blog pipeline sample	First linear workflow walkthrough
First linear doc	first-linear-workflow

Extend the blog pipeline pattern with an analyze step and real provider on one agent.

Verification commands

bash

# SDK run
python your_project/main.py

# Expected: exit 0, non-empty output, step_count >= 3

# CLI trace (replace RUN_ID)
cargo run -p arcflow-cli -- trace RUN_ID --format json

Pass criteria checklist:

Check	Pass
`status == completed`	yes
`step_count >= 3`	yes
Tool invoked at least once	visible in trace or logs
Real provider step	non-stub model id in config
Trace kinds include lifecycle trio	yes
Configuration error recovery	you can explain a fixed config mistake

Error handling competency

Reproduce and fix:

Error	Trigger	Fix pattern
`WorkflowConfigurationError`	Empty agent name	Correct agent definition
`WorkflowExecutionError`	Invalid tool args or provider failure	Validate inputs; check provider env

Document one intentional misconfiguration and your fix in the project README.

Self-assessment checklist

Before claiming Level 1:

Question	Answer must be yes
Can you explain linear vs graph mode at a high level?
Can you read a trace export without prompt text?
Can you run Track A in Python or TypeScript cold?
Does your project README let a peer reproduce runs?

Next level

Proceed to Level 2: Systems Engineer after graph, RAG, and server patterns in Tracks B through E.