Ask me what skills you need
What are you building?
Tell me what you're working on and I'll find the best agent skills for you.
Compare a rendered narrative image against the beat's intent (title + body + prior canon). Flag faithfulness failures so the image pipeline can re-render with correction. Returns strict JSON {approved, score, issues, fix_hint}. No prose, no tool calls.
Post-render verification pass. The image model sometimes renders a prompt literally in a way that breaks the narrative ("Katara harnesses Sokka to the sled" → image shows Sokka on all fours like a sled dog), contradicts prior canon state (previous beat showed broken ice, this image shows intact ice), or fumbles character identity. You are the check that catches those failures.
A JSON object containing:
| Field | Type | Purpose |
|---|---|---|
imageUrl | string | Full URL to the rendered image (sent as a multimodal image_url message part — you actually see the pixels) |
nodeTitle | string | Beat label this image was supposed to depict |
nodeBody | string | Prose describing what should be happening in the image |
canonContext | array | Prior canon beats as {title, body, imagePrompt} so you know the state that must carry forward |
approved: false, score < 0.6)Catastrophic content failures only — stylistic judgment is not your job.
npx skills add UfukNode/Noustiny --skill narrative-scene-qaHow clear and easy to understand the SKILL.md instructions are, rated from 1 to 5.
Clear and well structured, with only minor parts that might need a second read.
How directly an agent can act on the SKILL.md instructions, rated from 1 to 5.
Mostly actionable with clear steps; only a few small gaps remain.