What problem does it solve?
When working on complex, multi-step tasks that require dozens of tool calls across multiple sessions, the AI's volatile context window loses track of progress, past decisions, and research findings, leading to repeated work, inconsistent outputs, and broken context after session resets or /clear commands.
Core Features & Use Cases
- Persistent Planning Files: Creates three durable Markdown files (task_plan.md for phase tracking, findings.md for research and decisions, progress.md for session logs) that act as long-term working memory outside the context window.
- Automatic Session Recovery: Detects unsynced context from previous sessions after context resets, and provides a summary of missed work to resume tasks seamlessly.
- Built-in Guardrails: Enforces rules like the two-step operation rule to prevent loss of multimodal findings, error logging to avoid repeated failures, and a three-strike failure protocol to break out of unproductive loops.
- Use Cases: Ideal for product analysis, multi-step project planning, complex bug diagnosis, research tasks, and any workflow requiring more than 5 sequential tool calls.
Quick Start
Use the planning-with-files-zh skill to break down my upcoming 3-phase e-commerce checkout feature redesign into tracked steps and initialize the required planning files in my project directory.