DisclaimerUnofficial, and not affiliated with Anthropic. Nearly all of this is read straight out of what ships: npm bundles, captured prompts, published docs. Anthropic's own notes go in verbatim, marked as theirs. The rest is my reading, and every entry carries the strings behind it. If one looks wrong, vote it down and say why.
Sub-agents (e.g., Task agents) now receive an explicit instruction: "Do not emit text between tool calls. Inter-tool narration is never shown to the user — go straight to the next tool call." This reduces wasted tokens on text the user never sees.
Evidence
Instruction string A9Y (search for "Do not emit text between tool calls")
Strings lifted out of the shipped bundle, so the claim above can be checked against them.