Character Style Consistency By 1ronuwr

のぞむ＊AIイラスト · Character and style consistency

Turn the method into an executable, Case-grounded workflow. The Skill is incomplete unless the final artifact can be traced to one named anchor Case.

Inputs

Ask only for missing inputs:

• Intended subject, audience, and use.
• Target image model, aspect ratio, and output constraints.
• Required references, brand assets, copy, or identity details.
• Preferred anchor Case from references/cases.md, or permission to select one.

Required evidence gate

• Read references/cases.md before drafting on every invocation.
• Select exactly one anchor Case. Name it in the response; do not average several unrelated visual styles.
• Inspect the anchor's finished media with an available image, video, or browser tool. If the media cannot be inspected, say so and do not claim visual fidelity.
• Write a reference contract before the artifact:
  • Preserve: at least three structural traits supported by the Case evidence.
  • Replace: subject, copy, brand, or context supplied by the user.
  • Avoid: unrelated dominant motifs and superficial copying of creator identity.

Workflow

1. Restate the requested deliverable, audience, and success criteria.
2. Read references/cases.md and select one anchor Case.
3. Inspect its finished media, summary, and prompt excerpt.
4. Produce the Preserve / Replace / Avoid reference contract.
5. Fix identifying traits.
6. Add costume, props, and action.
7. Unify output with style and world constraints.
8. Produce the requested artifact using the output contract below.
9. Compare it with the anchor Case, then revise material failures once.

Output contract

• A reference contract naming one anchor Case, with Preserve, Replace, and Avoid decisions.
• A one-paragraph creative brief.
• A complete generation prompt with composition, subject, lighting, material, and style constraints.
• Negative constraints or likely failure modes.
• A short verification checklist.

Verification

• The prompt names one primary subject and a readable composition.
• Required text, identity, material, and aspect-ratio constraints are explicit.
• The verification checklist can be judged from the generated image.
• The response names one anchor Case and confirms that its evidence file was read.
• At least three visible or structural traits in the artifact map back to the Preserve list.
• No unrelated theme becomes dominant unless the user explicitly requested it.
• If finished media was inaccessible, the response labels the result as prompt-grounded rather than visually verified.
• Distinguish observed evidence from your own recommendation.
• Link or name the relevant GoodCase evidence used for the artifact.

Safety and attribution

• Do not invent an original prompt, model, result, creator, metric, or source.
• Do not imply this Skill is authored, endorsed, or officially distributed by a referenced creator.
• Preserve creator attribution when reproducing or discussing a source Case.
• Transfer structure and decision logic; do not impersonate a creator or present a close copy as original work.
• Do not publish, purchase, upload private assets, or make external account changes without explicit user approval.