Paper 1 · Question 13

A: --output-format json controls the output structure, not the analysis logic. It formats results as JSON for downstream parsing but does not affect the accuracy of the verification itself.

B: Without explicit validation criteria (e.g., 'for each exported function in src/api/, check that docs/api/ contains a section with the function name, parameters, and return type'), the model may perform a shallow check. Adding structured criteria, expected output format, and examples of what constitutes a gap produces reliable verification.

C: The -p flag runs Claude Code in non-interactive (pipe) mode. It does not restrict file access — Claude Code retains full access to its built-in tools (Read, Grep, Glob) for reading files.

D: Adding a second instance to review the first adds latency and cost without addressing the root cause. If the first instance's prompt lacks clear validation criteria, the second instance reviewing the same vague output will draw the same flawed conclusions.