FEFreeExamDumps.in

CCA-F Practice Questions — Page 14

Question 131

Open question ↗

Scenario: Customer Support Agent

The `process_refund` tool rejects a request because the order is 47 days old and outside the 30-day return window. This is a policy rule, not a temporary glitch, but the tool returns a generic failure and the agent keeps retrying with backoff.

How should this error be represented?

  • A.`errorCategory: "transient"` with `isRetryable: true` so the agent retries with backoff.
  • B.A generic `"Operation failed"` message with no metadata.
  • C.Return success with an empty result so the agent stops retrying.
  • D.`errorCategory: "business"` with `isRetryable: false` and a clear user-facing explanation.

Question 132

Open question ↗

Scenario: Customer Support Agent

A subagent that searches the knowledge base catches an internal timeout and returns an empty result set marked as success. The agent tells the customer "no relevant policy found" and closes the ticket, unaware the search actually failed.

What is the underlying anti-pattern?

  • A.Excessive escalation of routine errors to the coordinator.
  • B.Infinite retries inside the subagent.
  • C.Silent suppression: a failure was disguised as a valid empty result, so the agent cannot tell "no matches" from "search failed."
  • D.Overly narrow task decomposition by the coordinator.

Question 133

Open question ↗

Scenario: Customer Support Agent

Mid-conversation a customer writes: "Stop — I don't want a bot, connect me to a human agent now." The agent instead runs three more diagnostic tool calls trying to resolve the issue itself before eventually escalating.

What is the correct behavior?

  • A.Escalate to a human immediately, since an explicit request for a human is an immediate-escalation trigger.
  • B.Continue troubleshooting first; only escalate if the agent cannot resolve the issue.
  • C.Run a sentiment check to confirm the customer is genuinely frustrated before escalating.
  • D.Ask the customer to confirm twice that they really want a human.

Question 134

Open question ↗

Scenario: Customer Support Agent

To "future-proof" the support agent, the team gave it 17 tools spanning refunds, shipping, account edits, analytics, and marketing email sends. Tool-selection accuracy dropped from 94% to 78% versus an earlier build that had 5 role-relevant tools.

What is the most likely cause and fix?

  • A.The system prompt is too short; list all 17 tools with detailed descriptions to restore accuracy.
  • B.Too many tools degrade selection reliability; scope the agent to only the tools its role needs plus a few utilities.
  • C.Tool count is irrelevant; the regression must be a model-version change.
  • D.Switch `tool_choice` from `"auto"` to `"any"` to force more decisive selection.

Question 135

Open question ↗

Scenario: Customer Support Agent

The `lookup_order` tool returns 40+ fields per order, but resolving a ticket needs only `order_id`, `status`, `total`, `items`, and `return_eligible`. Across a long multi-order conversation the extra fields crowd the context window and the agent starts mixing up orders.

What is the most effective mitigation?

  • A.Tell the model in the system prompt to ignore the fields it doesn't need.
  • B.Switch to a model with a larger context window so all fields fit.
  • C.Call the tool less often and cache the full 40-field results for reuse.
  • D.Add a `PostToolUse` hook that trims each result to the five relevant fields before the model sees it.

Question 136

Open question ↗

Scenario: Developer Productivity Tools

In a Claude Code session you must (1) locate every file matching `**/*.config.json` and (2) find where a function named `resolveTenant` is defined. The agent tries to do both by reading directories one at a time, wasting turns.

Which built-in tools fit these two tasks?

  • A.Use `Glob` to find files by the name pattern, and `Grep` to search file contents for the function definition.
  • B.Use `Grep` to find files by the pattern, and `Glob` to search contents for the function.
  • C.Use `Read` on every directory to do both tasks.
  • D.Use `Bash` with `find` and `cat`, since no dedicated tools exist for this.

Question 137

Open question ↗

Scenario: Developer Productivity Tools

You must migrate an internal logging library used across 52 files in an unfamiliar service. Several refactoring approaches look viable and you don't yet know all the call sites or their coupling.

Which mode should you start in?

  • A.Direct execution: start editing files immediately and adjust as problems surface.
  • B.Planning mode: explore with Read/Grep/Glob, map dependencies, and produce a plan for approval before changing anything.
  • C.Direct execution with one long up-front instruction covering all 52 files at once.
  • D.Skip planning and run a single batch refactor command across the repo.

Question 138

Open question ↗

Scenario: Developer Productivity Tools

A multi-phase refactor needs an initial discovery pass that reads dozens of files and prints verbose output. Done in the main session, this discovery floods the context window and degrades later steps.

How should the discovery phase be run?

  • A.Run discovery inline in the main session and `/compact` afterward to recover space.
  • B.Raise `max_tokens` so the verbose discovery output fits comfortably.
  • C.Delegate discovery to an Explore subagent so its verbose output stays isolated from the main context.
  • D.Skip discovery and start editing, reading files only as errors appear.

Question 139

Open question ↗

Scenario: Developer Productivity Tools

During a multi-day investigation of a large codebase, the agent's later answers degrade — it begins referring to "typical service patterns" instead of the specific `TenantResolver` class it found on day one. You want findings to survive across context boundaries and new sessions.

Which technique helps most?

  • A.Write key findings (file paths, classes, dependencies, constraints) to a scratchpad file the agent consults later.
  • B.Keep the entire conversation in context and never compact it.
  • C.Ask the model to memorize the findings so it can recall them.
  • D.Restart each day from scratch with no notes to keep context clean.

Question 140

Open question ↗

Scenario: Developer Productivity Tools

You ask Claude Code to rename a local variable, but the `Edit` call fails because the target text appears three times in the file and the match is not unique.

What is the recommended fallback?

  • A.Retry the same `Edit` repeatedly until it eventually succeeds.
  • B.Delete the file and recreate it from memory with the rename applied.
  • C.Switch to `Bash` and use `sed` to force the replacement.
  • D.Use `Read` to load the full file, modify the content programmatically, then `Write` the updated version.