Implements a persistent file-based planning system to organize and track progress on complex, multi-step AI tasks.
This skill enhances Claude's reliability during long-running tasks by adopting the 'Manus' workflow, which treats the filesystem as persistent disk storage to supplement limited context window RAM. By creating and maintaining standardized markdown files—task_plan.md, findings.md, and progress.md—the agent ensures that critical project architecture, research discoveries, and session logs remain available even after context clears. It enforces rigorous protocols, such as the 2-Action Rule for saving data and a 3-strike error recovery system, making it essential for high-stakes development and deep research projects.
Key Features
01Standardized 3-strike protocol for advanced error handling and debugging
020 GitHub stars
03Structured templates for phase tracking and session logging
04Automated session recovery and context retention after clearing
05Persistent markdown-based project memory and state tracking
06The '2-Action Rule' for immediate research and discovery logging
Use Cases
01Managing complex multi-step software development projects
02Executing deep research tasks requiring extensive tool and browser calls
03Documenting architectural decisions and system migration steps
