qol-workflow-nodeslisted

# Workflow Nodes KMRH47 prefers repeated work to become deterministic scripts. The goal is not abstraction for its own sake. The goal is less human memory burden, fewer repeated agent tool calls, faster iterations, and artifacts that prove what happened. ## Default posture When work starts looking like a workflow, stop and ask: 1. What is the smallest human input? 2. What concrete output should exist after the run? 3. Which steps are deterministic enough to script? 4. Which steps still require human judgment? 5. What report would let the next human or agent continue without rediscovery? If the answer is clear, create or extend a workflow node instead of retyping the chain manually. ## When to script Script the workflow when at least one is true: - The same command chain is run more than once. - The workflow crosses tools, languages, VMs, games, devices, external CLIs, or generated artifacts. - A human has to remember flags, paths, order, cleanup, or scoring rules. - The agent would otherwise spend many tool calls discovering state each time. - The output can be validated by file existence, hashes, logs, exit codes, screenshots, reports, or structured data. - A later script could consume the result. Do not script a process that is still mostly unknown. Observe it once, identify the real inputs and outputs, then extract the repeatable core. ## Shape Use this layering: - **Leaf script**: one fragile domain operation. Examples: Blender export, asset conversion, table