Task Continuation Plugin

Minimal task continuation plugin for OpenCode - automatically continues sessions when incomplete tasks remain.

Overview

This plugin provides task continuation functionality:

1. Task Continuation: Automatically continues sessions when incomplete todos remain

Perfect for:

- Multi-step task execution with automatic continuation
- Long-running agent workflows
- Preventing premature session termination
- Ensuring all tasks in a todo list are completed

Installation

``bash
npm install @frugally3683/agent-loop-plugin
`

Usage

`typescript
import agentLoopPlugin from "@frugally3683/agent-loop-plugin"

export default agentLoopPlugin
`

Configuration

The plugin can be configured with the following options:

`typescript
{
countdownSeconds: 2, // Seconds to wait before auto-continuation
errorCooldownMs: 3000, // Cooldown period after errors
toastDurationMs: 900, // Toast notification duration
agent: "custom-agent", // Agent name for continuation prompts
model: "custom-model", // Model name for continuation prompts
logFilePath: "path/to/log", // Path to log file for debugging
}
`

Features

- Automatic Continuation: Sessions automatically continue when incomplete tasks remain
- User Message Handling: Cancels pending continuations when users send new messages
- Error Handling: Graceful handling of session errors with cooldown periods
- Recovery Support: Sessions can be marked as recovering to pause auto-continuation

API

$3

`typescript
import { agentLoopPlugin, createTaskContinuation } from "@frugally3683/agent-loop-plugin"
`

$3

`typescript
const continuation = createTaskContinuation(input, options)

continuation.handler({ event }) // Handle session events
continuation.cleanup() // Cleanup resources
continuation.scheduleContinuation(sessionID) // Manually schedule continuation
continuation.markRecovering(sessionID) // Mark session as recovering
continuation.markRecoveryComplete(sessionID) // Mark recovery as complete
continuation.cancel(sessionID) // Cancel pending continuation
``

License

MIT