feibot-ops

name	feibot-ops
description	Standardize lifecycle operations for the local feibot gateway service via launchd. Supports multiple instances, listing all instances, and provides guidance on allowFrom configuration.

Mybot Ops

Overview

Use this skill to run repeatable and safe lifecycle actions for feibot through one script. launchd is the primary supervisor (auto-start on login + keepalive). Supports multiple instances with different configurations.

Quick Start

/path/to/feibot-ops/scripts/manage.sh <command>

Commands:

• list - List all feibot instances (running and installed)
• start - Ensure launchd service is loaded and running
• stop - Stop launchd service
• restart - Restart gateway
• status - Show launchd/process/log status
• logs [N] - Tail last N log lines (default: 120)
• install - Write plist and bootstrap launchd service
• uninstall - Stop service, disable label, remove plist

CLI Options

-c, --config FILE    Config file path
-r, --repo DIR       Repository directory
-l, --label LABEL    Launchd label (must be unique per instance)
-d, --run-dir DIR    Runtime directory
-h, --help           Show help

Examples

# List all instances
./scripts/manage.sh list

# Default instance (auto-detected repo path + default config/run dirs)
./scripts/manage.sh restart

# Specify config file
./scripts/manage.sh -c /path/to/config.json status

# Different label for multi-instance
./scripts/manage.sh -l ai.feibot.prod -c /prod/config.json restart

# Full custom config
./scripts/manage.sh -r /path/to/feibot -c /path/to/config.json -l feibot-dev -d /path/to/run status

Creating a New Instance

When creating a new instance:

1. Create directory structure:

   mkdir -p ~/botname/workspace ~/botname/sessions ~/botname/run
   

2. Create config.json with:

   • Set name to the bot's name (e.g., "suzy", "zoe")
   • Configure Feishu channel (appId, appSecret)
   • Set the model in agents.defaults.model
   • Use allowFrom: [] for open access in testing, or a specific open_id list for restricted access (see below)

3. Create AGENTS.md in the workspace:

   # Agent Instructions
   
   You are a helpful AI assistant. Your name is <Bot Name>. Be concise, accurate, and friendly.
   
   ## Guidelines
   
   - Always explain what you're doing before taking actions
   - Ask for clarification when the request is ambiguous
   - Use tools to help accomplish tasks
   - `memory/MEMORY.md` is approved global context; `memory/HISTORY.md` is the searchable session index
   

4. Install and start:

   ./scripts/manage.sh -l ai.botname.gateway -c ~/botname/config.json -r /path/to/feibot -d ~/botname/run install
   

5. Configure allowFrom (see next section)

⚠️ Important: allowFrom Configuration

The allowFrom field in the Feishu channel config is special:

1. It should contain Feishu user open_id values.
2. Empty list [] means allow all.
3. For restricted mode, collect IDs from logs after users send messages. IDs appear like:

   Access denied for sender ou_xxxxxxxxxxxx on channel feishu
   

How to configure restricted allowFrom:

1. Start the bot and have the user send a message
2. Check the logs: tail -f /path/to/run/gateway.log | grep "Access denied"
3. Copy the ou_xxx ID from the log
4. Add it to the allowFrom list in config.json
5. Restart the bot

Example:

"allowFrom": [
  "ou_6287aa1fb258625bd15f8f9789a04799",
  "ou_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
]

Environment Contract

• Repository: $REPO_DIR
• Config: $CONFIG_FILE
• Runtime dir: $RUN_DIR
• Pid file: $RUN_DIR/gateway.pid
• Log file: $RUN_DIR/gateway.log
• launchd label: $LAUNCHD_LABEL
• launchd plist: ~/Library/LaunchAgents/$LAUNCHD_LABEL.plist

Rules

• Do not use pkill -9 unless normal stop flow fails.
• Each instance must have a unique LAUNCHD_LABEL.
• After restart, confirm both launchd state and process existence.
• Remind users to choose allowFrom mode explicitly: [] (open) or a restricted allow-list.

Output Contract

When responding to the user after running commands, always include:

1. action executed
2. resulting PID (if running)
3. launchd state
4. key log lines (startup or error)
5. evidence references: exact log file path + timestamp(s) used