[Feature Request] AssistantResponse hook for enforcing AI behavior standards

Problem

We want to catch lazy AI patterns like "let's skip this" or "do it manually" - but these appear in AI text responses, not tool calls.

Current hooks (PreToolUse/PostToolUse) only see tool I/O. Scanning tool output causes false positives - reading a file containing "skip" triggers the warning, even though the AI didn't decide to skip anything.

Example

1. AI runs: Bash("npm test") → 47 failures
2. AI responds: "Let's skip the failing tests for now."
3. AI moves on

Lazy behavior is in step 2. Hooks only fire on step 1.

Request

AssistantResponse hook with blocking capability:

{
  "hook_event_name": "AssistantResponse",
  "message_text": "Let's skip the failing tests for now."
}

Hook returns:

{ "decision": "block", "reason": "Detected 'skip'. Please address the issue." }

Use Cases

• Lazy pattern detection ("skip", "manually", "for now")
• Quality gates on AI reasoning
• Team standard enforcement

Related

See #17865 for observability angle. This focuses on enforcement - intervening, not just observing.