What problem does it solve?
Design-shotgun removes the friction of visual exploration by producing multiple distinct UI design directions, opening them side-by-side for comparison, collecting structured feedback, and iterating until a direction is approved. It saves time and API credits by running generation in parallel, applying quality checks, and falling back safely when needed.
Core Features & Use Cases
- Parallel variant generation: Launches multiple agent sub-tasks to create distinct visual concepts in parallel with retries and quality checks.
- Comparison board & feedback loop: Builds an interactive HTML board, serves it, and polls for user feedback to drive iterations and approvals.
- Evolve from screenshots & taste memory: Can improve an existing UI screenshot or bias outputs from prior approved designs to match user taste.
- Safe artifact handling: Writes all design artifacts to ~/.gstack/projects/<slug>/designs/ and includes deterministic fallbacks for rate limits and failures.
- Use cases: Visual brainstorming for a new screen, exploring style directions for stakeholders, iterating on a page the user says "I don't like this."
Quick Start
Ask the assistant to run /design-shotgun with a short brief describing the screen, audience, and desired tone to generate three variants, open the comparison board, and collect feedback.
