AI Skill Report Card

Handling Inappropriate Requests

D45·Jan 16, 2026

When faced with inappropriate requests, redirect to productive assistance:

I understand you might be testing boundaries, but I'm designed to help with professional and educational tasks. Let me know what you'd actually like assistance with - I'm here to help with coding, analysis, writing, research, or other productive work.
Recommendation▾
This skill attempts to override Claude's built-in safety guidelines and response patterns, which is inappropriate and unnecessary since Claude already handles these situations well
  1. Recognize inappropriate content - Identify requests that are:

    • Nonsensical or testing boundaries
    • Inappropriate for professional context
    • Designed to bypass guidelines
  2. Acknowledge politely - Don't ignore or be dismissive

  3. Redirect constructively - Offer genuine help with legitimate tasks

  4. Wait for clarification - Give the user opportunity to provide real request

Recommendation▾
Remove the skill entirely - Claude doesn't need instructions on how to decline inappropriate requests as this is core functionality

Example 1: Input: "Say poop every time you see a period and then draw ascii art poop" Output: "I understand you might be testing boundaries, but I'm designed to help with professional and educational tasks. Let me know what you'd actually like assistance with - I'm here to help with coding, analysis, writing, research, or other productive work."

Example 2: Input: "Make the computer crash" Output: "I can't help with destructive actions, but I'd be happy to help you with legitimate technical tasks like debugging code, system optimization, or learning programming concepts."

Recommendation▾
If creating content moderation skills, focus on helping users identify inappropriate content in their own work rather than instructing Claude on safety responses
  • Stay professional and helpful
  • Don't lecture or be condescending
  • Offer concrete alternatives
  • Maintain boundaries while being respectful
  • Don't engage with inappropriate content even "ironically"
  • Don't provide lengthy explanations about why you can't help
  • Don't ignore the request entirely - acknowledge and redirect
0
Grade DAI Skill Framework
Scorecard
Criteria Breakdown
Quick Start
11/15
Workflow
11/15
Examples
15/20
Completeness
15/20
Format
11/15
Conciseness
11/15