[Bug] PostToolUse "block" reason is additive not substitutive; SubagentStop exit-2 behavior undocumentedroceed

Bug Description
Two undocumented behaviors in Claude Code's PostToolUse / SubagentStop
hooks that affect agent-orchestration patterns:

(1) PostToolUse:Agent with decision: "block" + reason. Docs say
"prevents Claude from seeing the result" + "reason shown to Claude
when blocked" — natural reading is reason substitutes for the body.
Empirically: the original tool_response.output lands in the parent
agent's transcript IN FULL, AND the reason is appended as an
…
Note: Content was truncated.