hello is a preserved example: a single agent <Task> that runs the prompt in
.smithers/prompts/hello.mdx and greets whoever you name. It is no longer
installed by default; copy it from examples/init-pack/ when you want the
smallest end-to-end workflow.
name defaults to "world", so workflow run hello works on
its own.
Inputs
How It Runs
- Reads the input
name(defaults to"world"). - Runs one agent
<Task>whose prompt is.smithers/prompts/hello.mdx, withnameinterpolated into the Markdown via{props.name}. - Returns the agent’s greeting.
.smithers/prompts/hello.mdx (for example, ask for a
haiku instead of a greeting) and re-running.
Output Shape
hello returns the single task’s text output (the generated greeting). The run
log records the agent invocation and its result.
Use it when
- You are learning Smithers and want the smallest end-to-end run.
- You are scaffolding a new workflow and want a known-good single-task template to grow from.
Use something else when
- You need research, planning, and a validated implementation: use
implementorresearch-plan-implement. - You want to author a brand-new workflow interactively: use
create-workflow.
Resume & retry
Runs are durable and resume withup --resume. hello is a single task, so a
resume simply re-runs that task if it had not completed.