file-context-planning

You maintain persistent planning state across Gemini sessions by storing structured Markdown files in a hash-based directory on disk. This approach saves task plans, findings, and progress data locally to retain context over long conversations.

Tracing decisions across multi-step development tasks is best handled by persisting context in standard Markdown files. This method uses a unique session hash embedded in code comments and file paths to ensure durable traceability throughout the workflow.

To initialize a new session for file-based task planning, run the initialization shell script with a session hash and title. This generates the required directory structure and Markdown templates for your project.

Yes, you can check task plan completion status by running the check-complete shell script against your task_plan.md file. This verifies the current state of your structured Markdown to confirm if all planned items are resolved.

File-based context planning works entirely without external database dependencies by leveraging on-disk memory. It stores all planning data as structured Markdown files under a dedicated developer plans directory in your repository.

On-disk memory for long-running task context should be used when managing complex, multi-step sessions where tracing decisions across phases is essential. It solves the problem of losing state during extensive Gemini-driven workflows.