An AI engineer is developing a multimodal agent for a media company's archive team. The archive contains thousands of titles spanning two decades (full episodes, raw interviews, and unedited B-roll), ingested as monolithic files. Producers use the agent to answer visual content questions about archival footage, such as whether a specific asset contains a usable exterior shot of a landmark. When a producer asks this type of question, the agent's retrieval tool loads the full asset into context so the model can examine it, even though only a small portion of the asset is relevant. After a few such retrievals, the context window fills. The engineer needs to reduce context pressure through proactive design changes rather than relying on compaction.
Which proactive context management strategy should the engineer use to address this problem?

• A. Preprocess each archival asset into a scene-level metadata layer so the agent can identify the relevant segment from the metadata and retrieve only that portion for examination.
• B. Route each retrieved asset through a summarization step using a vision language model to produce a brief visual digest, passing only that digest into the agent's context window.
• C. Add explicit instructions to the agent's system prompt directing it to retrieve only the smallest necessary portion of each asset, so the model proactively limits what it loads.
• D. Configure the agent to invoke a text-only search of the catalog metadata before retrieving any footage, so it can answer from the catalog when possible without loading media.

A context engineer is reviewing an agent built on Databricks that helps warehouse operations staff with inventory queries. The agent is configured with a system prompt (120 tokens), a conversation history window (400 tokens), 10 tool definitions averaging 180 tokens each (1,800 tokens total), and a RAG pipeline that retrieves a 1,600-token document chunk at runtime. The total context regularly exceeds the model's 4,096- token limit, causing conversation history to be truncated and degrading multi-turn coherence. Usage logs show low session variety, repetitive task patterns, and an average of 3-4 tools invoked per session.
Which change to the agent configuration will most effectively reduce the token budget and improve model focus?

• A. Reduce the conversation history window from 400 tokens to 100 tokens and redistribute the free space across the tool definitions and retrieval chunk equally.
• B. Reduce the RAG retrieval chunk size from 1,600 tokens to 800 tokens, halving the retrieval footprint and freeing space for additional tool definitions.
• C. Implement a dynamic tool-routing layer that inspects each user turn and injects only the tools relevant to the detected intent, removing unused definitions from the context window.
• D. Replace the 10 tool definitions with a static curated set of the 4 most frequently invoked tools, reducing tool token cost to approximately 720 tokens per turn.

An AI engineer is auditing a Databricks compliance agent that has handled 65 conversation turns for a quarterly financial audit. The context window is at 92% capacity, and several audit findings remain open. The engineer wants to recover space by clearing specific raw tool outputs from the message history without compacting the conversation, and is reviewing four tool_result blocks. They also want to minimize the risk of losing information the agent still relies on.
The history contains:
A ledger reconciliation output at turn 7 returning 8,800 tokens of raw transaction rows from a read-only general ledger. The agent calculated aggregated taxable and non-taxable totals from these rows and passed them to the tax classification call at turn 10. The agent's later replies and tool calls do not quote any underlying transaction rows.
A tax classification output at turn 10 returning 5,200 tokens of per-line bucket breakdowns across 312 transaction lines. The agent extracted the bucket totals and passed them to the compliance flag call at turn 13.
The agent's later replies and tool calls do not quote any per-line bucket assignments.
A compliance flag output at turn 13 returning 95 tokens with three flag codes and one-sentence explanations.
The agent quotes specific flag codes and explanations from this output in turns 24, 38, and 61, incorporating them into the audit findings it is currently drafting.
A vendor master lookup output at turn 19 returning 3,400 tokens of supplier records and tax IDs from a read- only vendor master database. The agent quotes a vendor name and tax ID from this output in turns 41 and 57.
Which tool outputs should the engineer clear?

• A. The ledger reconciliation (turn 7) and tax classification (turn 10) outputs, because the agent's later reasoning relies on the compliance flag summary (turn 13) rather than on the upstream raw data.
• B. No outputs from the chain (turns 7, 10, 13), because the agent may need to re-quote raw transaction rows or per-line bucket assignments in the open audit findings still being drafted.
• C. The ledger reconciliation (turn 7), tax classification (turn 10), and vendor master lookup (turn 19) outputs, because they are the three largest blocks and yield the most context capacity at once.
• D. Only the tax classification output (turn 10), because the ledger reconciliation is the root of the dependency chain and must be retained for re-derivation if downstream conclusions are challenged.

A context engineer is building a retail agent with Agent Bricks Custom Agents. The agent has two registered tools: a SQL query tool that queries a live inventory database, and a vector search tool that searches a static product catalog vector index for product descriptions, specifications, and categories. Both tools are configured correctly, function properly, and return valid results when called.
At runtime, when users ask questions such as "Is this item in stock?", the agent consistently calls the product catalog tool and provides a plausible availability answer without checking the live inventory database. The product catalog tool is not intended to answer live stock availability questions.
Which context engineering issue explains this behavior?

• A. The tool descriptions and routing guidance are too ambiguous to distinguish live inventory availability questions from static product catalog search.
• B. The product catalog vector index was built from a periodic snapshot, so it may contain stale product metadata.
• C. The inventory tool's description does not explicitly mention "stock," "in stock," or "availability," even though it correctly queries real-time inventory when called.
• D. The agent prompt does not require source-of-truth checks before generating a final answer, so the agent treats any plausible retrieved context as sufficient.