planning-with-files

You can manage complex multi-step tasks by persisting planning data to disk using structured files like task_plan.md, findings.md, and progress.md. This external memory approach reduces cognitive load and prevents context drift across many tool calls.

File-based planning is a deterministic workflow that maintains a predefined phase schema on disk. It automatically creates task_plan.md, findings.md, and progress.md in your project directory to guide iterative decisions and capture learnings during research or software development.

To start planning a project, create a task_plan.md file in your project directory and follow the provided templates to organize your work. Helper scripts are available to initialize sessions, track progress across phases, and stop tasks with clear status and error capture.

Yes, file-based external memory is ideal for data analysis workflows with many tool calls. It implements a plan-driven reboot mechanism that persists findings and progress to disk, ensuring iterative decision making remains structured across long, complex analytical sessions.

The best way to track task status and capture errors is using a predefined phase schema with helper scripts that start, track, and stop tasks. This deterministic workflow automatically records status updates and errors directly into progress.md files on disk.

On-disk planning files are unnecessary for simple, single-step tasks that do not benefit from external memory or iterative decision making. If your workflow requires few tool calls and minimal phases, the overhead of maintaining task_plan.md and progress.md may be redundant.