@supierior/experimental-workflows

Pi Workflower package for cool experimental workflows that are useful as playgrounds for engine patterns, even when they do not have a practical product purpose.

This package depends on @supierior/workflower and registers experimental workflows as workflow-only skills and commands.

It currently registers these workflows:

counter
counter-loop
stateful-grilling
stateful-grilling-finalize

counter and counter-loop

Workflow ids: counter, counter-loop

Use these tiny workflows to play with Workflower handoffs, garden state, and loop-style repetition through skills.

The counter workflow initializes the active garden state key counter from user-provided integer values, then hands off to counter-loop by calling Workflower's workflower_handoff tool. The counter-loop workflow reads that garden state, increments current, saves the updated counter value, and either calls workflower_handoff for another counter-loop flower when current < end or stops when current >= end.

counter-loop is an internal handoff workflow: it does not register a user-facing /wf:counter-loop command. Users start /wf:counter; the loop is entered and repeated through workflower_handoff. The counter workflows use workflower_state_get and workflower_state_set instead of output files or pollen.

During the active garden, the state is stored at:

.workflower/workflows/<garden>/state.json

Smoke test

/wf:counter demo-counter

Enter a starting value and ending value when prompted. After the counter garden state is saved, run:

/next

The loop handoff and later loop iterations are configured to auto-advance.

stateful-grilling

Workflow ids: stateful-grilling, stateful-grilling-finalize

Use this workflow to demonstrate a state-cleared interview loop. The public stateful-grilling workflow asks the user 1-3 focused feature-discovery questions, then updates the active garden state key statefulGrilling.feature with a durable feature understanding and an understandingPercent estimate. If the estimate is below 95%, the workflow hands off to another stateful-grilling flower so the next mini-interview starts with cleared context and only the garden state as durable memory.

When the estimate reaches 95% or higher, the workflow hands off to the hidden stateful-grilling-finalize workflow, which writes the final artifact:

.workflower/workflows/<garden>/<sequence>-stateful-grilling-finalize/feature-description.md

The final workflow sets cleanupOnCompletion: false, so the final document remains after completion while loop flowers can be cleaned up.

Smoke test

/wf:stateful-grilling demo-feature

Answer the mini-interview questions. When ready for the workflow to consolidate that mini-conversation, run:

/next

The update step auto-advances into routing. Routing uses Workflower handoff tools for loops or finalization, and the final document generation step auto-completes.

Useful Workflower commands

/wf status
/wf stop
/wf list