endurance

What problem does it solve?

This Skill addresses the challenge of executing autonomous agents for extremely long-running, complex tasks that would otherwise exhaust standard context windows, leading to performance degradation and increased costs.

Core Features & Use Cases

• Context Engineering: Optimizes context window usage by prioritizing quality over quantity and employing summarization techniques.
• State Externalization: Manages agent memory and state in external files (e.g., WORKING_NOTES.md, STATE.json) rather than solely relying on chat history.
• Auto-Prompting Workflows: Breaks down large tasks into a chain of distinct, auto-generated prompts, enabling stable and intelligent execution over extended periods.
• Use Case: A multi-day code refactoring task where the agent needs to maintain context and progress across numerous stages without losing track of previous decisions or incurring excessive token costs.

Quick Start

Use the endurance skill to manage a long-running agent process by externalizing its state and auto-prompting the next steps.