An agent checks inventory at the start of a 20-turn session. 15 turns — Claude Certified Architect – Foundations Practice — Claude Cert Academy

An agent checks inventory at the start of a 20-turn session. 15 turns later it books stock based on the original check. The booking fails — stock was sold in the meantime. What anti-pattern caused this?

• The agent should have used a larger context window to remember the original data more reliably
• Stale context: volatile data (inventory) was fetched once and used much later without re-validation
• The inventory tool had insufficient caching
• The booking tool lacked a retry mechanism

Answer: Stale context: volatile data (inventory) was fetched once and used much later without re-validation

For volatile data like inventory, prices, or service status, agents must re-fetch immediately before acting — not rely on data fetched earlier in a long session. Stale context is a common agentic bug. Context window size, tool caching, and retries do not address the fundamental issue of using outdated data for consequential decisions.

Continue to Claude Cert Academy