菜单
Use when a BioChemInsight vision model needs to classify candidate structure images, reject Markush or fragment noise, or extract a compound ID from a highlighted structure image.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
Use when a BioChemInsight language model needs to normalize assay markdown into JSON or extract a final compound ID from free-form textual reasoning or OCR output.
Shared runtime and schema conventions for all BioChemInsight language and vision model calls, including default system prompts, retry policy, temperature, and reusable output-contract patterns.
基于 SOC 职业分类
| name | biocheminsight-vision-models |
| description | Use when a BioChemInsight vision model needs to classify candidate structure images, reject Markush or fragment noise, or extract a compound ID from a highlighted structure image. |
| model | inherit |
| effort | high |
| context | auto |
Use this skill for runtime vision-model tasks.
This skill inherits shared runtime conventions from biocheminsight-model-common and only overrides vision-task-specific behavior.
strict, balanced, or permissive.value, unit, and symbol/grade description.decisions item per provided page.COMPOUND_ID="None" only with ID_SOURCE="none" and evidence explaining why no reliable ID is visible.references/classify_structure_candidate_prompt.mdreferences/classify_structure_crop_check_prompt.mdreferences/classify_structure_border_review_prompt.mdreferences/detect_structure_pages_prompt.mdreferences/review_structure_pages_prompt.mdreferences/structure_to_id_prompt.mdreferences/review_structure_primary_id_prompt.mdreferences/runtime.jsonreferences/output_schemas.json