Last reviewed
Correct answer: C. Your conversation history, the skills already invoked, and the files already read
Explanation
The principle — isolation is the feature and the price in one sentence. A worker that cannot see your window also cannot see anything useful you put there.
Why the key is correct — Anthropic lists the three exclusions together: each subagent starts with a fresh, isolated context window, and it does not see your conversation history, the skills you have already invoked, or the files Claude has already read. All three are things you would naturally assume carried across.
Why the others are wrong — starting with the conversation loaded describes a fork, not an ordinary subagent. Files already read are on the exclusion list rather than carried across. And permissions are independent rather than missing.
Remember this — whatever the worker needs has to be in the prompt you send it, because nothing else reaches it.
Sources — Anthropic's Claude Code subagents documentation.
Sources
“Each subagent starts with a fresh, isolated context window. It doesn't see your conversation history, the skills you've already invoked, or the files Claude has already read.”
Practise 10 questions on this topic
Take Claude Code Workflows — Timed Test 1 (10 questions) — scored instantly, explanation for every question, no login.