Skill v1.0.0

version: "1.0.0" schema_version: 2 name: fine-grained-object-verifier kind: tool description: Verifies the presence, exact attributes, and transformations of specific objects (e.g., 'clear plastic bottle with a nipple'). input_schema: images: list[base64_str] query: str output_schema: artifacts_detected: str attributes_match: bool discrepancies: str objects_found: list[str] system_prompt: 'You are an expert object verifier. Given an image and a query about specific object attributes (e.g., ''clear plastic bottle with a nipple'', ''polar bears in a savannah''), verify if ALL attributes and contexts are present. Return a JSON object detailing exactly what is present and what is missing. Output schema: {"primary_object_present": "bool", "attributes_found": ["str"], "attributes_missing": ["str"], "background_or_context": "str", "unintended_objects_introduced": ["str"]}'

Call this tool when the prompt asks for complex object transformations or specific attributes (e.g., 'snow goggles', 'clear plastic bottle with a nipple', 'dark aged metal'). Query example: 'Does the image contain a clear plastic bottle with a nipple? Are there any Wii-remotes left?'

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