Feature request: PostToolUse hook support for SendMessage (inter-agent communication logging)

Problem

When using agent teams, there's no way to observe or log inter-agent communication. SendMessage is not in the list of tools that fire PostToolUse hooks.

Currently hookable tools: Bash, Edit, Write, Read, Glob, Grep, Agent, WebFetch, WebSearch, AskUserQuestion, ExitPlanMode, MCP tools.

SendMessage is missing from this list.

Use Case

We run a team of 7 specialized agents (analysts, engineers, PM, architect) that communicate via SendMessage. We want to:

1. Audit trail: Log all inter-agent messages to a shared file (like a team Slack channel)
2. Observability: See what agents are telling each other without reading each agent's full output
3. Debugging: When an agent acts on bad advice from another agent, trace the communication chain
4. Cross-team synthesis: A "solutions engineer" agent reviews the chat log to find patterns

Current Workaround

We instruct every agent to manually append messages to a team/chat.md file before calling SendMessage. This works but is fragile — agents sometimes forget, and the format isn't guaranteed.

Proposed Solution

Add SendMessage to the list of tools that fire PreToolUse and PostToolUse hooks. The hook input should include:

{
  "tool_name": "SendMessage",
  "tool_input": {
    "to": "agent-name-or-id",
    "message": "the message content"
  }
}

This would allow a simple hook like:

{
  "event": "PostToolUse",
  "if": { "tool_name": "SendMessage" },
  "handler": {
    "type": "command",
    "command": "bash -c 'cat /dev/stdin | jq -r \"\\\"**\\(.tool_input.from // \\\"Agent\\\")**: @\\(.tool_input.to) \\(.tool_input.message)\\\"\" >> team/chat.md'"
  }
}

Environment

• Claude Code CLI
• Using agent teams with 7 custom agents defined in .claude/agents/
• Agents actively communicate via SendMessage during analysis and implementation workflows