task-decomposer

To generate a task plan, you provide an immutable intent and accumulated state to the system, which then outputs a validated plan.json file containing sequential, actionable steps for automated systems or manual agent operations.

Intent decomposition handles state management by tracking established facts, known gaps, and dead ends within the state, ensuring that the generated task plan remains traceable and accounts for missing evidence during execution.

The best way to break down a high-level intention is to use an automated task planner that creates a validated plan.json schema, ensuring every step has traceable requirements and verifiable outcomes for your specific workflow.

Yes, you need Python support and file operation permissions to save and execute tasks, because the system requires an execution context to run the driver that processes each task and records the results.

Yes, you can use this task planning approach for replacing a live system without downtime by decomposing the high-level intention into a detailed, sequential plan.json that an AI agent can execute step by step.

A limitation of using a plan.json schema is that it requires an explicit execution context and a dedicated driver to execute each task and record results, making it less suitable for environments without file operation permissions.