Creative Qc

Workflow

1. Identify the artifact type:
   • brief
   • storyboard
   • prompt pack
   • generated images
   • generated videos
2. Run structural checks:
   • language consistency
   • count alignment
   • one-to-one mapping between scenes and prompts
   • hook strength and opening variety
   • CTA presence when the content is promotional
   • platform fit such as 9:16, 16:9, short-form pacing, or still-image use
3. Run creative checks:
   • visual coherence
   • tone consistency
   • prompt specificity
   • repeated or generic phrasing
4. If real assets are provided, compare them against the plan:
   • ImageUnderstandingAgent for images
   • VideoUnderstandingExpert for videos
5. Return prioritized issues and concrete fixes.
6. Only rewrite automatically if the user asks for direct revision. Otherwise review first.

Severity Rules

• pass: no material blockers, only optional polish
• revise: usable direction but there are issues that should be fixed before generation or handoff
• block: high-risk mismatch, missing inputs, or structural failure that should stop the next step

Handoff Examples

If a prompt pack needs source comparison first:

{"input_path":"inbox/reference.png","mode":"prompt"}

If a generated video needs intent comparison:

{"input_path":"inbox/output.mp4","mode":"shot_breakdown"}

If the QC result says the script itself should be revised:

{"input_text":"<script to revise>","mode":"rewrite","constraints":"fix repetition and keep the original meaning"}

Output Contract

Return a QC summary like this:

{
  "status": "pass",
  "critical_issues": [],
  "major_issues": [],
  "minor_issues": [],
  "fix_plan": [
    "..."
  ],
  "residual_risks": [
    "..."
  ]
}

Guardrails

• Distinguish objective mismatch from subjective taste.
• Review first; do not silently rewrite the user's material unless asked.
• If there are no issues, say what you checked so the pass result is meaningful.
• When reviewing prompts, keep model compatibility in mind: concise, unambiguous, and medium-appropriate prompts usually beat overloaded prompts.
• Prioritize findings that would waste generation budget, break platform fit, or create scene-to-prompt mismatch.