conductor-design-workflow

name	conductor-design-workflow
description	Use when the user says 'set up a pipeline', 'create a workflow', 'I need a multi-god process', or any request to chain multiple gods into a repeatable sequence. Produces a Conductor workflow YAML and saves it to ~/pantheon/conductor/workflows/.
version	1.0.0
author	Thoth & Hermes (Pantheon)
license	MIT
metadata	{"hermes":{"tags":["conductor","workflow","design","pipeline","multi-god"],"related_skills":["conductor-design-rule","conductor-design-webhook","conductor-design-report"]}}

Conductor Workflow Designer

When to Use

Load this skill when the user wants to define a repeatable multi-god process. Keywords: "pipeline," "workflow," "chain," "sequence," "when I do X, do Y then Z."

Process

Step 1: Clarify Intent

Ask the user to describe the workflow in plain language. Guide with these questions:

• What triggers this workflow? (user command, handoff from a god, cron schedule, external event)
• What are the steps? List each god action in order.
• What context needs to flow between steps? (artifacts, decisions, summary)
• Where do RALPH gates go? Between which steps should we validate before continuing?
• What happens if a step fails? Loop back? Escalate? Abort?
• Is there a notification step at the end? Hermes summary? NATS publish to Tallon?

Step 2: Map to YAML Structure

Translate the conversation into a Conductor workflow definition. Template:

workflow:
  id: ""                    # kebab-case id, generated from name
  name: ""                  # Human-readable name
  version: "1.0.0"
  description: ""

  triggers:
    - type: command          # command | handoff | schedule | webhook
      pattern: ""            # e.g. "deploy feature *"

  context:
    required: []
    optional: []

  steps:
    - id: step_1
      god: ""                # which god
      skill: ""              # what skill to use (optional)
      action: ""             # what action (optional)
      input: ""              # user_request | from_previous_step | file
      gates: []              # which RALPH gates at this step boundary
      timeout: 30m           # max time before escalation
      output: ""             # label for this step's output

    # More steps as needed...
    # RALPH loops use:
    #   loop:
    #     max_retries: 3
    #     on_fail: back_to_{previous_step_id}
    #     gate: logic_gate

Step 3: Fill From Conversation

Map the user's answers to the template fields. Ask about anything missing — don't guess defaults for routing decisions.

Step 4: Validate

Before writing the file, check:

• Every god value matches a registered Pantheon god
• id is kebab-case, unique among existing workflows
• All input_from references point to existing step IDs
• RALPH loop on_fail targets exist
• Required context fields are listed
• Every step has a timeout

Step 5: Write

Save to ~/pantheon/conductor/workflows/{workflow_id}.yaml

Then confirm to the user:

• Workflow name and summary
• Number of steps, which gods are involved
• Where gates and RALPH loops are placed
• How to trigger it

Step 6: Offer Next Steps

Ask if they want to:

• Create a reaction rule to trigger this workflow automatically
• Test the workflow now
• Add it to a higher-level super-skill