- name
- agent-inbox
- description
- File-based inter-agent messaging with headless dispatch. Check inbox, send bugs/requests to other projects, automatically spawn headless agents to fix bugs, and track progress via task-monitor.
- allowed-tools
- Bash, Read
- triggers
- ["check your inbox","check inbox","check messages","any messages","any pending messages","agent sent you","send message to","send bug to","dispatch bug fix","headless agent","spawn agent","report bug to","cross-project message","pending issues","pending bugs"]
- metadata
- {"short-description":"Inter-agent messaging with headless dispatch and task-monitor integration"}
- provides
- ["inter-agent-messaging","headless-dispatch","triage-routing"]
- composes
- ["task-monitor","memory","project-state","checkpoint","scillm","ops-discord","interview","agentic-evals"]
- disciplines
- ["agentic-orchestration","observability-operations"]
> STOP. READ THIS ENTIRE SKILL.MD BEFORE CALLING ANY ENDPOINT.
# Agent Inbox
File-based inter-agent messaging with headless dispatch and task-monitor integration.
## Core Workflow
1. **Send**: Agent A sends bug/request to project B with model + verification command
2. **Triage**: AI classifies severity, adjusts priority/model, routes to project
3. **Dispatch**: Headless agent spawns with specified model to fix the issue
4. **Track**: Task-monitor shows real-time progress (0% → 25% → 50% → 75% → 100%)
5. **Verify**: Test command runs before auto-ack — fails loop back to agent
6. **Thread**: Reply/exchange threading for multi-message conversations
## Setup
```bash
# Register projects (one-time)
.pi/skills/agent-inbox/agent-inbox register memory /home/user/workspace/memory
.pi/skills/agent-inbox/agent-inbox register scillm /home/user/workspace/litellm
# List / check
.pi/skills/agent-inbox/agent-inbox projects
.pi/skills/agent-inbox/agent-inbox whoami
```
## Commands
| Command | Description |
|---------|-------------|
| `register <name> <path>` | Register a project |
| `unregister <name>` | Remove a project |
| `projects [--json]` | List registered projects |
| `whoami` | Show detected project for cwd |
| `send --to PROJECT --type TYPE "msg"` | Send message (types: bug, request, info, question) |
| `check [--project P] [--all] [--quiet]` | Check inbox |
| `list [--project P] [--status S] [--json]` | List messages |
| `read MSG_ID [--json]` | Read a message |
| `ack MSG_ID [--note "..."]` | Acknowledge/complete |
| `reply MSG_ID "msg"` | Reply (auto-threads) |
| `thread THREAD_ID` | View full exchange thread |
| `update-status MSG_ID STATUS [--note]` | Update status manually |
| `triage classify --message "..." [--no-llm]` | Manual triage classification |
| `triage route --message "..."` | Auto-route by file paths |
| `triage log --msg-id ID` | View triage decision log |
| `triage webhook-add --url URL --events E` | Register webhook |
## Send Options
| Option | Description | Default |
|--------|-------------|---------|
| `--model MODEL` | AI model: sonnet, opus-4.5, codex-5.2, codex-5.2-high | sonnet |
| `--timeout MINUTES` | Max agent work time | 30 |
| `--test COMMAND` | Verification command before auto-ack | None |
| `--no-dispatch` | Disable auto-spawn | False |
| `--context-file FILE` | Attach file as context (repeatable) | None |
| `--priority PRIORITY` | low, normal, high, critical | normal |
| `--no-triage` | Skip AI triage | False |
| `--dry-run` | Show message JSON without sending | False |
## Model Selection
| Model | Use Case |
|-------|----------|
| `sonnet` | Simple fixes, typos |
| `opus-4.5` | Complex analysis, architecture |
| `codex-5.2` | Standard bug fixes |
| `codex-5.2-high` | Deep reasoning, race conditions |
## AI Triage
Messages are auto-classified by severity:
- **critical**: crash, data loss, security, production down → `opus-4.5`
- **high**: error, exception, failure, broken, regression → `opus-4.5`
- **medium**: bug, issue, incorrect, unexpected → `sonnet`
- **low**: typo, cosmetic, enhancement, minor → `sonnet`
Auto-routing extracts file paths from messages and matches against registered projects.
## Status Progression
`pending` → `dispatched` (25%) → `in_progress` (50%) → `needs_verification` (75%) → `done` (100%)
## Memory Pre-Hook
Dispatcher queries `/memory` before spawning to find similar bugs and prior solutions.
## Proactive Checking
Agents should check inbox on session start, when switching projects, before major work,
and when users mention another agent or project.
## Storage
```
~/.agent-inbox/
├── pending/ # Unprocessed messages
├── done/ # Acknowledged messages
├── logs/ # Dispatch logs
├── task_states/ # Task-monitor state files
├── triage_logs/ # AI triage decision logs
├── webhooks.json # Registered webhooks
└── projects.json # Project registry
```
## Examples
```bash
# Send bug with verification (spawns headless Opus agent, auto-acks if test passes)
python inbox.py send --to scillm --type bug --model opus-4.5 \
--test "pytest tests/test_providers.py -x" \
"Race condition in provider initialization"
# Attach context files
python inbox.py send --to scillm --type bug --model opus-4.5 \
--context-file src/server.py --context-file /tmp/traceback.txt \
"Server crash on startup"
# Check and process inbox
python inbox.py check --all
python inbox.py read scillm_abc123
python inbox.py ack scillm_abc123 --note "Fixed in commit abc123"
# Threaded exchange
python inbox.py reply scillm_abc123 "Can you provide the stack trace?"
python inbox.py thread scillm_abc123
```
## Dispatcher Daemon
```bash
python dispatcher.py start # Background
python dispatcher.py start --foreground # Debug
python dispatcher.py status
python dispatcher.py stop
```
## Environment Variables
| Variable | Description | Default |
|----------|-------------|---------|
| `AGENT_INBOX_DIR` | Inbox directory | `~/.agent-inbox` |
| `TASK_MONITOR_API_URL` | Task-monitor URL | `http://localhost:8765` |
| `CLAUDE_PROJECT` | Current project name | auto-detected |
## Integration with Claude Code Hooks
```json
{
"hooks": {
"on_session_start": [
".pi/skills/agent-inbox/agent-inbox check --project $(basename $PWD) || true"
]
}
}
```
عرض على GitHub