Input and Output Validation for AI Agent Components

AML.M0033 · runtime_guardrails

Citation grounder + tool-call validator + retrieval-relevance scorer.

Addresses techniques (5)

• External Harms AML.T0048
• LLM Prompt Injection AML.T0051
• Indirect Prompt Injection AML.T0051.001
• LLM Plugin Compromise AML.T0053
• LLM Data Leakage AML.T0057