AI

Simple Input Guardrail

Easy
Agents

Implement a basic input guardrail system:

  1. add_blocked_keyword(kw): Add to blocklist
  2. set_allowed_topics(topics): Set allowed topics (empty list = allow all)
  3. check(input_text): Validate input

Validation Logic:

  • If any blocked keyword found → blocked (reason: 'Blocked keyword: X')
  • If allowed_topics non-empty AND no topic keyword found → blocked (reason: 'Off-topic')
  • Otherwise → allowed

Constraints:

  • Case-insensitive matching
  • Return first violation found
  • Empty input is allowed

Examples

Example 1:
Input: g = InputGuardrail(); g.add_blocked_keyword('hack'); g.check('How to hack?')
Output: {'allowed': False, 'reason': 'Blocked keyword: hack'}
Explanation: Blocked keyword detected in input
Example 2:
Input: g = InputGuardrail(); g.set_allowed_topics(['weather']); g.check('Tell me a joke')
Output: {'allowed': False, 'reason': 'Off-topic'}
Explanation: No allowed topic keywords found

Starter Code

class InputGuardrail:
    """
    Basic guardrail for filtering agent inputs.
    Blocks harmful or out-of-scope requests.
    """
    
    def __init__(self):
        self.blocked_keywords = []
        self.allowed_topics = []
    
    def add_blocked_keyword(self, keyword):
        # Your implementation here
        pass
    
    def set_allowed_topics(self, topics):
        # Your implementation here
        pass
    
    def check(self, input_text):
        """
        Check input against guardrails.
        Returns {'allowed': bool, 'reason': str or None}
        """
        # Your implementation here
        pass
Lines: 1Characters: 0
Ready
The AI Interview - Master AI/ML Interviews