CCAR-P Advanced Agentic Architecture Practice Question
An agent is engaged in a multi-hour troubleshooting session involving dozens of tool calls and thousands of lines of log data. The architect notices that the agent is starting to 'forget' early symptoms of the problem. Which strategy best addresses this while managing token costs?
⚠ Common exam trap
Candidates frequently suggest simply increasing the context window or dumping all logs into the prompt. This ignores the exponential cost of token usage and the degradation of model focus over time.
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
✓
Implementing a recursive summarization buffer
Managing context in long-running agentic sessions requires a balance between detail and capacity. A summarization strategy combined with a sliding window allows the agent to retain the 'gist' of historical turns while keeping the most recent, high-fidelity data available. This prevents context overflow while maintaining the logical continuity of the troubleshooting process.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Truncating the oldest messages once the limit is reached
Why it's wrong here
Simple truncation removes the earliest parts of the conversation, which often contain the critical initial problem statement or symptoms. Without that context, the agent may lose sight of the original goal or repeat failed troubleshooting steps that were already attempted at the beginning of the long session.
- ✓
Implementing a recursive summarization buffer
Why this is correct
Summarizing previous turns into a concise narrative preserves the essential findings and progress of the agent. By passing this summary forward into new turns, the agent maintains a continuous understanding of the session history without needing to re-process every individual token from the original, verbose tool outputs.
- ✗
Switching to a model with a 1-million token window
Why it's wrong here
While a larger window exists, using it for every turn in a long session exponentially increases latency and cost. Furthermore, models can still experience 'lost in the middle' phenomena where they struggle to attend to specific details buried deep within a massive context, making this a poor architectural choice.
- ✗
Storing all tool results in an external vector database
Why it's wrong here
Vector databases are excellent for retrieval-augmented generation but are less effective for maintaining the exact sequential logic of a troubleshooting flow. Relying solely on RAG might cause the agent to pull in disjointed facts without understanding the temporal relationship between the different steps it has already taken.
About these practice questions
Courseiva writes every CCAR-P question from scratch — 262 in total, each with an explanation and a wrong-answer breakdown. None are copied from real exams or 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.