CCA-F: Claude Certified Architect — Foundations

By Anthropic · Question Mode

QuestionQ45

Agentic Architecture & Orchestration

Compliance requires that refunds exceeding $500 automatically escalate to a human agent; this rule cannot be left to model discretion. Despite clear system-prompt instructions, production logs show that the agent occasionally processes high-value refunds directly, with a 3% failure rate. How can guaranteed compliance be achieved?

Explanation

A deterministic pre-execution hook enforces the policy outside the model’s discretion: it can inspect the requested refund amount, block any amount above $500, and invoke the required human-escalation path. Prompt wording and examples cannot provide a guarantee because model behavior is not a hard authorization boundary.

Community Discussion

No comments yet. Be the first to start the discussion!