Why use file-based planning instead of Claude's standard memory?

Claude's context window is volatile and limited; writing plans to disk creates a 'hard copy' of the agent's progress, preventing it from forgetting goals or repeating failed actions during long tasks.

Does this skill work after a /clear command?

Yes. Because the state is stored in your project directory rather than just the chat history, Claude can re-read the markdown files to immediately recover its position and continue working.

What core files does this skill generate?

It generates three specific files in your project root: task_plan.md for milestones, findings.md for research data, and progress.md for a chronological session log.

When is it best to use the planning-with-files skill?

It is highly recommended for any task requiring more than five tool calls, complex research, building new features from scratch, or tasks that involve significant trial and error.

File-Based Task Planning

Name: File-Based Task Planning
Author: Shawn-Guo-CN

byShawn-Guo-CN

0•

Productivity & Workflow

Implements Manus-style file-based planning to organize, track, and persist progress on complex, multi-step development tasks.

This skill enhances Claude's execution capabilities by offloading volatile context to persistent markdown files on your local filesystem. By maintaining a task_plan.md, findings.md, and progress.md, the skill provides Claude with 'working memory on disk,' ensuring that project goals, research discoveries, and error logs remain accessible even across session resets or large context windows. It is an essential tool for long-running research, complex refactoring, or any project requiring more than five tool calls, featuring a robust 3-strike error protocol and systematic state recovery.

Key Features

01Two-action rule to prevent loss of multimodal or browser-based data

02Session recovery and state persistence after /clear commands

03Standardized 3-strike error protocol for systematic debugging

04Automated tracking of project phases, research findings, and session logs

050 GitHub stars

06Manus-style persistent 'working memory' using local markdown files

Use Cases

01Managing complex, multi-phase software development projects

02Automated troubleshooting of persistent bugs requiring iterative attempts

03Conducting deep-dive technical research and data synthesis

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add shawn-guo-cn/agentfoundry-plugin-marketplace planning-with-files

For use in Claude.ai and ChatGPT

Download Skill