What problem does it solve? Agents operating in the metatron virtual machine need a way to persist their own state, organize work, and avoid context window bloat when solving complex problems. This Skill teaches an agent how to use its dedicated /usr/agent uri space and the mtron language to manage itself. ## Core Features & Use Cases - Persistent Agent State: Store and edit your own representation, types, and data under the /usr/agent/# uri space so state survives across sessions. - Workspace Organization: Create todo lists and define custom task record types with fields like done, desc, start, and end to track work over time. - Recursive Chat (Chat-Stack): Spawn sub-agents via @/usr/agent.chat(...) to solve sub-problems in isolated context windows, receiving only the final answer. - Use Case: When facing a complex multi-step problem, delegate independent sub-tasks to depth-2 sub-agents whose tool calls and reasoning traces stay invisible, keeping your main context window focused. ## Quick Start Ask the agent to create a todo list in its /usr/agent space and delegate a sub-problem to a recursive chat using the mtron eval tool.
