Design Agent Droid

Run a focused UI polish or mockup-to-implementation pass through Factory/Droid using the custom droid `design-agent-droid`. Prefer GLM-5V Turbo when available; current working fallback/default is GLM-

SKILL.md

GitHub

Design Agent Droid

Use this skill when you, the orchestrating agent, need to launch Factory/Droid for a visual-first UI pass.

What this skill is for

This is an orchestrator-facing launch skill. It tells the current agent how to route the task, shape the prompt, and choose the right Droid-side worker.

Use it to:

  • choose the correct Droid agent/droid
  • shape the delegation prompt for visual/design work
  • avoid generic coding-agent prompting when the task is really UI/UX polish

Do not tell the downstream Droid worker to load, activate, or read this skill itself unless the user explicitly wants that. The normal flow is:

  1. You read this skill.
  2. You launch Droid correctly.
  3. The Droid worker does the task using its own droid instructions.

Default workflow

  1. Inspect the current UI first.
  2. Name the design intent before editing.
  3. Make the smallest coherent improvement pass.
  4. Verify desktop/mobile and important interaction states.
  5. Report what changed, why it helped, and what still needs another pass.

Preferred agent

On Factory/Droid, prefer the custom droid:

  • design-agent-droid

Model guidance

  • GLM-5V Turbo remains the preferred visual model family when accessible.
  • On this setup, the current working fallback/default is GLM-5.1 via the Z.ai coding endpoint.
  • Do not block the task just because GLM-5V is quota- or plan-gated.

Recommended prompt shape

"Use design-agent-droid for a UI polish pass on <surface>. Inspect the current UI or screenshots first, state the design intent, make the smallest high-leverage improvements, then report visual QA findings, interaction QA findings, and remaining weaknesses."

Output checklist

  • Design intent summary
  • Top UI/UX issues found
  • Concrete changes recommended or applied
  • Breakpoint and interaction QA notes
  • Remaining risks / next pass