<< All versions

Skill v1.0.0

currentAutomated scan100/100
tiger-ai-lab/rewardharness/embrace-intended-drastic-changes
──Details
PublishedNovember 3, 2026 at 08:48 AM
Content Hashsha256:ccdfe8a3c1ebb757...
Git SHA0c1e7051ab4f
──Files
Files (1 file, 1.1 KB)
SKILL.md1.1 KBactive
SKILL.md · 20 lines · 1.1 KB

schema_version: 2 name: embrace-intended-drastic-changes kind: skill description: Rules for evaluating setting and style changes, strictly forbidding penalizing an image for losing original context when a new setting is requested. metadata: version: "1.0.0"


Embrace Intended Drastic Changes

When a prompt requests a completely new setting (e.g., 'cafeteria', 'child's bedroom', 'grassy savannah', 'underground subway'), you MUST expect and reward the introduction of new, context-appropriate elements (e.g., beds, grass, fluorescent lights, subway tiles).

CRITICAL RULE: DO NOT penalize an image for 'disrupting the original context' or 'failing to preserve the original background' when a setting change is explicitly requested.

An image that preserves the original setting (e.g., keeping a classroom when a cafeteria is requested, or keeping a restaurant background) FAILS the instruction. Drastic changes are the goal. Reward the image that most convincingly establishes the NEW requested setting, even if it means replacing the entire background.

<!-- Release review: RewardHarness v0.4.1 (2026-08-26). -->

All versions