<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> This Claude Code skill implements a persistent markdown planning workflow inspired by Manus AI, the company acquired by Meta for $2 billion. It addresses common AI agent limitations like volatile memory and goal drift by leveraging markdown files as a structured "working memory" on disk. The skill transforms complex task execution for Claude Code users, enabling robust planning, progress tracking, and knowledge storage, mirroring the pattern attributed to Manus AI's rapid success.

How It Works

The core approach employs a "3-File Pattern": task_plan.md for progress, notes.md for research, and [deliverable].md for output. This pattern acts as the AI's persistent memory, overcoming context window limitations. An iterative loop ensures the AI re-reads the task_plan.md before key decisions, maintaining focus and preventing goal drift, a key "context engineering" principle from Manus AI. This filesystem-centric memory management allows for complex, multi-step task execution without losing track of objectives.

Quick Start & Requirements

Installation involves cloning the repository directly into your Claude Code skills directory (e.g., ~/.claude/skills/). The primary requirement is a functional Claude Code environment. No other non-default prerequisites like specific hardware, CUDA versions, or large datasets are mentioned. Setup is minimal, primarily involving a git clone command.

Highlighted Details

• Implements Manus AI's context engineering principles: filesystem as memory, attention manipulation, error persistence, goal tracking, and append-only context.
• Automatically generates task_plan.md and notes.md for complex tasks, updating progress via checkboxes in the plan file.
• Logs failures within the plan file for iterative improvement and re-reads the plan before critical decision points.
• Designed for multi-step tasks, research, project building, and workflows requiring extensive tool calls.

Maintenance & Community

Contributions are welcomed via standard pull request workflows. While the README acknowledges Manus AI for pioneering the pattern and Anthropic for the Claude Code framework, it does not provide direct links to community channels (like Discord/Slack), a roadmap, or specific maintainer contact information.

Licensing & Compatibility

The project is released under the permissive MIT License. This license allows for free use, modification, and distribution, making it highly compatible with commercial applications and closed-source projects without significant restrictions.

Limitations & Caveats

This skill is specifically tailored for Claude Code and is intended for multi-step, complex, or research-heavy tasks. It is not designed for simple queries, quick lookups, or single-file editing operations where its overhead would be unnecessary.