CCAR-P Advanced Agentic Architecture Practice Question
You operate a long-running research agent that maintains a scratchpad of findings across many turns. You notice that as the scratchpad grows, the agent begins ignoring recent tool results and repeating earlier conclusions. You cannot increase the context window. Which intervention most directly addresses the root cause of the recency failure?
⚠ Common exam trap
The trap here is diagnosing the recency failure as a memory or determinism problem and reaching for external storage or temperature changes, when the actual cause is where recent evidence sits inside a growing context.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Restructure the scratchpad into a fixed-size rolling summary that is regenerated each turn, with the most recent tool results kept verbatim in a dedicated, clearly delimited section.
The agent is not forgetting recent results so much as losing them in an ever-growing block of text where early material dominates attention. Bounding the scratchpad with a regenerated rolling summary keeps total size stable, and reserving a clearly delimited section for the newest tool results puts current evidence in a position the model reliably attends to. This fixes the structural cause rather than the sampling or storage symptoms.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Move the scratchpad out of the prompt entirely and store it in an external vector database, retrieving relevant entries only when the agent explicitly asks.
Why it's wrong here
Externalizing the scratchpad helps with scale but replaces it with a retrieval problem: the agent must now formulate a query that surfaces the newest evidence, which it may fail to do precisely because it is already ignoring recent signals. The recency failure is about what sits in the active window, so removing the scratchpad entirely from the prompt does not guarantee the latest results are in view.
- ✗
Append every tool result verbatim to the scratchpad so no information is lost, and instruct the agent to re-read the entire scratchpad before each decision.
Why it's wrong here
Appending verbatim makes the scratchpad grow faster, which is the opposite of what is needed, and instructing a full re-read consumes even more of the fixed context window. The agent's tendency to weight earlier material over recent material is a positional effect, not a recall deficit, so adding more text at the end does not fix it and likely worsens the dilution.
- ✓
Restructure the scratchpad into a fixed-size rolling summary that is regenerated each turn, with the most recent tool results kept verbatim in a dedicated, clearly delimited section.
Why this is correct
The root cause is that accumulated history pushes recent results into a diluted middle position. A rolling summary compresses old findings into a bounded block, while a dedicated recent-results section guarantees the newest evidence sits in a high-attention position. This keeps total context roughly constant, so recency is preserved without needing a larger window.
- ✗
Lower the model's temperature to zero so the agent becomes more deterministic and less likely to drift from the current evidence.
Why it's wrong here
Temperature controls sampling randomness at generation time; it does not change how the model attends to positions in a long input. The failure is a context-position effect, not a sampling-variance effect, so a deterministic decode will still ignore recent tool results if they sit in a diluted part of the window. This treats a symptom while leaving the structural cause untouched.
About these practice questions
This CCAR-P question is part of Courseiva's 262-question bank — original exam-style content with full explanations and wrong-answer analysis, never real exam questions or exam dumps. Learn why practice questions differ from exam dumps →
JA
Written and reviewed by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
Last reviewed September 2026 · checked against the official Anthropic exam blueprint
This CCAR-P practice question is part of Courseiva's free Anthropic certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the CCAR-P exam.