Implements persistent, file-based task planning and research tracking to handle complex multi-step AI workflows.
Planning with Files is a specialized skill that enhances Claude's capabilities by moving its 'working memory' from the volatile context window to persistent markdown files on disk. Inspired by Manus-style agents, it creates a structured environment using task_plan.md, findings.md, and progress.md to track goals, research, and session history. This approach is essential for long-running tasks where context window limits might otherwise lead to lost information, ensuring the AI maintains focus, follows a rigid 3-strike error recovery protocol, and never repeats failed actions during complex migrations or research projects.
Key Features
01Automated plan verification hooks that check progress before critical tool executions
02Persistent working memory using standardized markdown files (task_plan.md, findings.md, progress.md)
032-Action rule for immediate persistence of research and multimodal findings
041 GitHub stars
05Manus-style planning logic optimized for complex, multi-step AI tool chains
06Standardized 3-strike error protocol to ensure intelligent troubleshooting and escalation
Use Cases
01Executing large-scale software migrations or complex architectural refactors
02Conducting deep research projects requiring multiple web searches and data synthesis
03Managing development tasks that span dozens of tool calls and require state recovery
