prompt-quality-auditor

name	prompt-quality-auditor
description	Audit prompt systems for enforceable product and UI quality. Use when the user asks for a deep prompt review, prompt quality audit, UI-generation prompt hardening, or review of prompts for layering, shadows, glow, animation, app chrome, accessibility, responsive behavior, state coverage, or consistency across prompt files.

Prompt Quality Auditor

Use this skill to evaluate whether prompts reliably enforce high-quality output, not merely whether they mention good design words.

Workflow

1. Inventory prompt sources.

   • Locate prompt files, schema/formatter code, allowed lists, templates, and runtime prompt assembly.
   • Identify which prompts control concept generation, style selection, variation, artifact generation, and critique.
   • Check for generated runtime copies when the app uses a prompt asset sync step.

2. Audit enforcement, not vocabulary.

   • Look for required baselines for layering, z-index, depth, shadows, glow, pseudo-elements, motion, reduced-motion behavior, app chrome, responsiveness, accessibility, and UI states.
   • Check whether requirements are explicit, repeated in the right prompt stages, and represented in schemas or formatters.
   • Flag contradictions, vague adjectives, invalid examples, and prompts that ask for generation when strict selection is required.

3. Compare prompt files against each other.

   • Identify uneven standards between artifact prompts, style prompts, design templates, variation prompts, and concept prompts.
   • Confirm that downstream prompts preserve quality requirements introduced upstream.
   • Verify that allowed lists and examples match the actual runtime options.

4. Produce findings.

   • Prioritize systemic gaps over wording preferences.
   • Include specific file references, observed weakness, likely output failure, and recommended prompt/schema change.
   • If app code must change to preserve prompt intent, state that separately from prompt-copy changes.

5. Recommend hardening order.

   • Start with the artifact/output prompt baseline.
   • Normalize strict selection prompts next.
   • Add shared requirements to variation and concept prompts.
   • Update schemas, formatters, tests, and prompt asset sync if they enforce the contract.

Guardrails

• Do not count aesthetic terms as evidence of enforceable quality.
• Do not recommend UI prompt changes without checking how prompts are assembled at runtime.
• Do not make UI code changes unless the user explicitly asks for implementation.