AgentSkillsCN

regulatory-impact-assessor

在根据监管要求评估AI项目时使用。建议在部署前使用。该技能可生成合规性评估报告、差距分析，以及相应的整改计划。

SKILL.md
--- frontmatter
name: regulatory-impact-assessor
description: Use when assessing AI projects against regulatory requirements. Use before deployment. Produces compliance assessment, gap analysis, and remediation plan.

Regulatory Impact Assessor

Overview

Assess AI projects against applicable regulatory requirements. Identify compliance gaps, document risk levels, and plan remediation to ensure lawful and ethical AI deployment.

Core principle: Regulatory compliance is not optional. Proactive assessment prevents costly remediation and enforcement actions.

When to Use

  • •Before AI project approval
  • •Before production deployment
  • •When regulations change
  • •Annual compliance review
  • •After significant model changes

Output Format

yaml
regulatory_assessment:
  project: "[Project name]"
  assessment_date: "[YYYY-MM-DD]"
  assessor: "[Name/Team]"
  
  system_profile:
    description: "[What the AI does]"
    decision_type: "[Support | Automation | Recommendation]"
    affected_parties: "[Who is impacted]"
    data_types: ["[Data type]"]
    deployment_regions: ["[Region]"]
  
  applicable_regulations:
    - regulation: "[Regulation name]"
      jurisdiction: "[Where applies]"
      applicability: "[Why this applies]"
  
  requirements_analysis:
    - regulation: "[Regulation]"
      requirements:
        - requirement_id: "[REQ-001]"
          requirement: "[Requirement text]"
          status: "[Compliant | Partial | Gap | N/A]"
          evidence: "[How demonstrated]"
          risk: "[High | Medium | Low]"
  
  gap_summary:
    critical:
      - gap: "[Gap description]"
        remediation: "[How to fix]"
        deadline: "[When needed]"
    
  remediation_plan:
    - action: "[Action]"
      owner: "[Who]"
      timeline: "[When]"
  
  approval:
    ready_for_deployment: "[Yes | No | Conditional]"
    conditions: ["[If conditional]"]

Key Regulations

EU AI Act

yaml
eu_ai_act:
  risk_categories:
    prohibited:
      - "Social scoring"
      - "Manipulation of vulnerable groups"
    
    high_risk:
      - "Employment decisions"
      - "Credit decisions"
      - "Education access"
      requirements:
        - "Risk management system"
        - "Data governance"
        - "Technical documentation"
        - "Human oversight"
    
    limited_risk:
      - "Chatbots"
      requirements:
        - "Transparency (disclose AI use)"

Assessment Checklist

  • • System profile documented
  • • Applicable regulations identified
  • • Requirements mapped
  • • Compliance assessed
  • • Gaps documented
  • • Remediation planned
  • • Approval obtained