CCARF Practice Question
Correct answer: B
A: Even with /compact, each service's context (source code, documentation, edits) is substantial. Sequential processing risks context degradation as compaction summaries lose detail from earlier services, and it is slower than parallel processing.
B: git worktree creates separate working directories on different branches, each with its own isolated Claude Code session. Each session has a full context budget dedicated to one service. The three updates run in parallel without interfering with each other, and results are merged via standard git workflows.
C: context: fork isolates a skill's output from the main conversation but does not create multiple parallel execution contexts. A single skill cannot simultaneously process three independent services in parallel.
D: The batch API is for high-throughput, latency-tolerant workloads — not for interactive Claude Code sessions. Documentation updates require interactive exploration of source code and iterative editing, which the batch API does not support.