Trinity Remote Operations

Collaborate with remote Trinity agents - execute prompts, sync-and-run, and configure notifications.

Command Overview

Arguments

$ARGUMENTS

Agent Name Detection

This skill automatically detects the agent name using these methods (in order):

1. •

   template.yaml (preferred):

   bash

   grep "^name:" template.yaml 2>/dev/null | cut -d: -f2 | tr -d ' '
   

2. •

   Directory name (fallback):

   bash

   basename "$(pwd)"
   

3. •

   Environment variable (override):

   bash

   echo "$TRINITY_AGENT_NAME"
   

Status Check (default)

When called without arguments, show remote agent status.

Workflow

1. •

   Determine agent name (see detection methods above)

2. •

   Query remote status

   code

   mcp__trinity__list_agents()
   mcp__trinity__get_agent(name: "<agent-name>")
   

3. •

   Display status

   code

   ## Remote Agent Status
   
   | Property | Value |
   |----------|-------|
   | Name | <agent-name> |
   | Status | running |
   | Uptime | 3h 24m |
   | Last activity | 2 minutes ago |
   | Git branch | main |
   | Git HEAD | 1cb2c9a |
   

Remote Execution (exec)

Execute a prompt on the remote agent without syncing local changes.

Use when: You want to run something on the always-on remote version without pushing local work.

Workflow

1. •

   Verify remote agent exists and is running

   code

   mcp__trinity__list_agents()
   

   If not running:

   code

   mcp__trinity__start_agent(name: "<agent-name>")
   

2. •

   Extract prompt from arguments (everything after "exec ")

3. •

   Execute on remote

   code

   mcp__trinity__chat_with_agent(
     agent_name: "<agent-name>",
     message: "<the prompt>"
   )
   

4. •

   Return results directly to user

Example

User: /trinity-remote exec check my calendar for tomorrow

Agent:
→ Remote agent: my-agent
→ Status: running
→ Executing...

[Response from remote agent appears here]

Deploy and Run (run)

Sync local changes to remote, then execute a task.

Use when: You've made local changes (skills, CLAUDE.md, etc.) and want to test them on the remote agent.

Workflow

1. •

   Check for uncommitted changes

   bash

   git status --porcelain
   

2. •

   If changes exist, commit and push

   bash

   # Stage skill/policy/procedure changes
   git add .claude/skills/ .claude/agents/ .claude/rules/ CLAUDE.md memory/
   
   # Commit with auto-message
   git commit -m "[deploy-run] Auto-sync before remote execution"
   
   # Push to origin
   git push origin $(git branch --show-current)
   

3. •

   Tell remote to pull

   code

   mcp__trinity__chat_with_agent(
     agent_name: "<agent-name>",
     message: "Run: git fetch origin && git pull origin main --ff-only"
   )
   

   Wait for confirmation.

4. •

   Execute the task Extract prompt (everything after "run "):

   code

   mcp__trinity__chat_with_agent(
     agent_name: "<agent-name>",
     message: "<the prompt>"
   )
   

5. •

   Return results with sync summary

   code

   ## Deploy-Run Complete
   
   ### Sync
   - Committed: 3 files
   - Pushed to: origin/main
   - Remote pulled: ✓
   
   ### Execution
   [Response from remote agent]
   

Example

User: /trinity-remote run generate the weekly analytics report

Agent:
→ Local changes detected: 2 files
→ Committing and pushing...
→ Remote pulling...
→ Executing task on remote...

## Weekly Analytics Report
[Full response from remote]

Notification Configuration (notify)

Configure how to receive notifications from remote executions.

Sub-commands

Storage

Store notification config in memory/trinity_notifications.json:

{
  "agent_name": "<agent-name>",
  "notifications": {
    "email": "user@example.com",
    "webhook": null,
    "enabled": true
  },
  "subscriptions": [
    {
      "trigger": "schedule_complete",
      "notify": ["email"]
    },
    {
      "trigger": "error",
      "notify": ["email", "webhook"]
    }
  ]
}

Workflow for notify status

1. •Read memory/trinity_notifications.json
2. •Display current configuration:

## Notification Status

| Setting | Value |
|---------|-------|
| Email | user@example.com |
| Webhook | (not configured) |
| Enabled | ✓ |

### Active Subscriptions
- Schedule completions → Email
- Errors → Email

Workflow for notify email <address>

1. •Validate email format
2. •Update memory/trinity_notifications.json
3. •Confirm:

✓ Email notifications set to: user@example.com

You'll receive notifications for:
- Scheduled task completions
- Execution errors
- Agent status changes

Workflow for notify webhook <url>

1. •Validate URL format
2. •Test webhook with a ping:
   bash

   curl -X POST "<url>" -H "Content-Type: application/json" \
     -d '{"type": "test", "agent": "<agent-name>", "message": "Webhook configured"}'
   

3. •Update config if successful
4. •Confirm:

✓ Webhook configured: https://your-endpoint.com/notify
✓ Test ping successful

Webhook payload format:
{
  "type": "schedule_complete|error|status_change",
  "agent": "agent-name",
  "task": "task description",
  "result": "success|failure",
  "message": "details...",
  "timestamp": "ISO-8601"
}

Quick Reference

Error Handling

Related Skills

• •trinity-adopt - Initial agent setup
• •trinity-sync - Manual git sync operations
• •trinity-schedules - Scheduled task management
• •trinity-compatibility - Compatibility audit