proactive-tasks_skill

Overview

This skill is a proactive goal and task management system that turns reactive assistants into autonomous partners. It tracks goals, breaks them into actionable tasks, and enables agents to work during heartbeats, message humans when needed, and make steady progress without waiting for prompts. The system includes reliability patterns to prevent data loss and recover from context truncation.

How this skill works

The skill stores goals and tasks in structured JSON and provides a CLI for adding, querying, and updating items. Every critical operation writes a WAL entry, updates a SESSION-STATE file, and appends to a working buffer to survive context cuts. During heartbeats or isolated agent turns it picks the next work-ready task, works for a short interval, logs progress, and notifies humans only for meaningful events or blockers.

When to use it

• When managing multi-step projects or long-term goals
• When you want an agent to autonomously make incremental progress during heartbeats
• When you need robust recovery from context truncation or interruptions
• When tasks require verification before reporting completion
• When you want structured human notifications for completions or blockers

Best practices

• Break goals into specific, single-session tasks (15–60 minutes) with clear completion criteria
• Enable heartbeat or isolated cron jobs for autonomous work; avoid using systemEvent for background tasks
• Rely on WAL → SESSION-STATE → working-buffer flow to prevent lost updates and enable recovery
• Run health-check periodically and after recoveries to detect and auto-fix common inconsistencies
• Verify outcomes end-to-end before marking tasks completed (Verify Before Reporting)

Example use cases

• Autonomously research and compare hardware or models, logging findings and next steps
• Run recurring maintenance tasks during heartbeats and report only major findings or blockers
• Track development milestones and verify features before marking them complete
• Recover work after context truncation using working-buffer, SESSION-STATE, and WAL logs
• Manage exploratory projects by converting scope discoveries into new subtasks and dependencies