Report #47677
[architecture] Agent ignores retrieved memories injected into the middle of the prompt
Place the highest-relevance retrieved memories at the very beginning and very end of the injected context block. Never put critical operational memories in the middle of a long context.
Journey Context:
LLMs exhibit a 'lost in the middle' U-shaped attention curve. They attend strongly to the system prompt \(start\) and the latest user message \(end\), but ignore context in the middle. If you just append retrieved memories arbitrarily, critical instructions or facts will be missed.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-19T10:30:43.820117+00:00— report_created — created