What problem does it solve? Running an AI agent continuously across sessions requires stitching together scheduling, memory persistence, remote dispatch, and computer control, which standalone frameworks like Hermes or AutoGPT handle as monolithic systems. This Skill replaces those frameworks by composing Claude Code's native crons, dispatch, memory, and MCP servers into a self-directing agent loop. ## Core Features & Use Cases - Persistent Memory: Combines built-in markdown memory files with the MCP memory server's knowledge graph for cross-session recall of entities, relations, and observations. - Scheduled Operations: Creates cron-based recurring tasks such as daily PR reviews, hourly health monitors, and nightly test runs using cron expressions. - Dispatch and Computer Use: Triggers remote agents from CI/CD or webhooks and automates browsers and desktops via the computer-use MCP server. - Task Queue: Maintains a memory-persisted task list that survives session boundaries. - Use Case: Set up an autonomous PR reviewer that checks watched repositories every 30 minutes, runs tests, reviews changes, posts comments via the GitHub MCP, and records status in memory. ## Quick Start Ask the agent to create a scheduled task that reviews open pull requests every weekday at 9 AM and writes a summary to memory.
