Parallel workflow agents stagger their first call to share cache; tune with an env var.
What's wrong with this entry?
When several workflow agents start with an identical prompt prefix, the followers now wait briefly for the leader's first response so their request hits the prompt cache instead of paying full price for the same prefix. The wait is capped by CLAUDE_CODE_WORKFLOW_PREFIX_STAGGER_MS, which has a built-in default when unset and is forced to 0 when DISABLE_PROMPT_CACHING is set.
- Agents sharing a prefix register in a warm-up map; the first one through is the leader
- Followers wait up to the cap, then proceed regardless
- With prompt caching disabled there is no wait at all
CLAUDE_CODE_WORKFLOW_PREFIX_STAGGER_MS
Strings lifted out of the shipped bundle, so the claim above can be checked against them.