design-agent-automation-workflowlisted

# Design Agent Automation Workflow Design agent and automation workflows before implementation. This skill is a framework-neutral planning surface. It helps choose the smallest defensible automation shape, name the safety and state boundaries, decide whether the task is fit for full automation or needs exact escalation gates, and produce a scaffold another stack-owned skill or implementation pass can use. ## Inputs - Required: the automation goal or workflow idea - Useful: target repository, cadence, write surface, expected outputs, full-auto eligibility, escalation points, state needs, retry needs, observability needs, and preferred runtime constraints - Optional: known framework preference, deployment target, language stack, or existing scheduler/service ## Workflow 1. Restate the intended real-world outcome and the smallest useful first run. 2. Decide whether automation is appropriate yet. Prefer safe full automation when bounded scope, validation, rollback or draft behavior, and side-effect controls make it reasonably reliable. Use human review only for the exact decision that cannot be made safe through narrower scope, deterministic checks, retries, rollback, sandboxing, or an orchestration layer. 3. Classify the best-fit surface: - Codex app automation - `codex exec` or Codex GitHub Action - Codex subagent fan-out - OpenAI Agents SDK service - LangGraph graph - Hermes-specific workflow - full-auto execution - auto-with-esc