Discuss Anthropic CCAR-F Exam Topic 3 Question 4

Your pipeline reviews every pull request using a single API call with a static prompt containing the diff and the full text of each changed file; unchanged files are not included. Reviews are posted asynchronously and do not block pull-request creation. Developers report that reviews consistently miss bugs involving cross-file interactions---for example, a pull request renames a function's parameters, but the review does not flag callers in other files that still use the old parameter names. Post-release analysis shows that cross-file bugs account for 35% of production incidents from reviewed pull requests. What is the most effective change to your review design?

ARedesign the review as a turn-limited agentic task in which the model can read files and search the codebase through tools, following references to verify cross-file findings.

BAdd chain-of-thought instructions asking the model to list all external references in the diff and then reason step by step about how each change might affect callers in other files.

CUse static analysis to build a dependency graph of changed code, and then expand the prompt to include every file within two dependency hops of any changed file.

DRun parallel review passes for each changed file with its direct dependents included, and then aggregate and deduplicate the findings through a final summarization call.

Suggested Answer: A

The failure is caused by missing evidence, not insufficient reasoning over the supplied evidence. A static prompt containing only changed files cannot reliably identify callers, configuration dependencies, generated interfaces, or indirect relationships located elsewhere in the repository. Asking Claude to reason more deeply cannot recover code that it was never given.

Option A converts the review into a bounded agentic loop. Claude can search for symbol references, read relevant callers, inspect type definitions, and follow newly discovered dependencies before validating a potential defect. Anthropic describes the Claude Code agentic loop as gathering context, taking action, verifying results, and repeating based on tool feedback. A turn limit preserves predictable cost and execution time.

Option C improves coverage but imposes an arbitrary two-hop boundary and depends on the accuracy of a precomputed graph, which may omit dynamic imports, reflection, generated code, configuration references, or language-specific call relationships. Option D creates duplicated context and aggregation complexity while still limiting each reviewer to predetermined dependents. Option B changes the reasoning instructions but supplies no mechanism for checking unchanged files. Tool-enabled, just-in-time retrieval is therefore the most adaptable and reliable architecture for cross-file review.

================

Get Premium CCAR-F Questions as Interactive Web-Based Practice Test or PDF Real Exam Format Detailed Explanations Instant Access Get Your Practice Test

12 comments · newest first

Colby 27 days ago

Option B sounds good too. Step-by-step reasoning might help identify issues better.

Roxanne 1 month ago

I prefer option C. A dependency graph could catch more cross-file bugs.

Barb 1 month ago

I think option A is the best. It allows the model to actively search for issues.

Natalie 1 month ago

A seems like it would give the model more context to work with.

Lilli 2 months ago

Not sure if B would be enough to catch all the issues.

Nickole 2 months ago

Wow, 35%? That's a huge number!

Terrilyn 2 months ago

I think D is overkill, just stick with A for simplicity.

Keena 2 months ago

Sounds like option C could really help with those cross-file bugs!

Tammi 2 months ago

I feel like option A could be the most effective since it allows for more dynamic interaction with the codebase, but I’m not clear on how it would be implemented in practice.

Shawn 2 months ago

Chain-of-thought reasoning in option B sounds interesting, but I’m not confident it would catch all cross-file issues since it relies on the model's reasoning ability.

Renay 3 months ago

I think option D could work well since it aggregates findings, but I wonder if running multiple passes might slow down the review process too much.

Tien 3 months ago

I remember we discussed how static analysis can help identify dependencies, so option C seems promising, but I'm not entirely sure if it covers all edge cases.

Anthropic CCAR-F Exam - Topic 3 Question 4 Discussion