菜单
Produce structured JudgeRubric outputs for workflow quality gates.
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。
基于 SOC 职业分类
Use when designing, implementing, configuring, testing, or extending applications built with @purista/harness and its provider adapters, including agents, workflows, tools, skills, models, state, sandbox, telemetry, and custom adapter packages.
Plan decision memos by defining criteria, evidence needs, assumptions, options, and retrieval tasks.
Convert research structure into Mermaid or draw.io-ready artifact instructions.
Review drafts for unsupported claims, contradictions, missing alternatives, weak citations, and bias.
Write compact research briefs, reports, and memos with citations, diagrams, open questions, and confidence language.
Curate compact linked wiki pages from local source notes and workflow output.
| name | judge-rubric |
| description | Produce structured JudgeRubric outputs for workflow quality gates. |
Score workflow outputs with criteria that match the task. Use approved only when evidence is sufficient and risks are addressed. Use needs_human_review, revise, or rejected when the result requires human judgement or more work.
Return machine-readable rubric fields with criterion ids, labels, scores, and rationales.
