What is the '2-Action Rule' in this skill?

It requires the agent to save findings to a file after every two search or browser operations to prevent multimodal or visual information from being lost when the context window shifts.

How does the 3-Strike Error Protocol work?

It guides the agent through three distinct phases: diagnosis/fix, alternative approach, and a broader rethink of assumptions before finally escalating the issue to the user.

When should I use this skill versus standard Claude interaction?

This skill is best for tasks requiring more than 5 tool calls, complex research, or multi-phase projects where you need a persistent audit trail and objective tracking.

Where are the planning files stored?

While templates and scripts are stored in the skill directory, your active planning files (task_plan.md, findings.md, and progress.md) are created in your project's root directory.

Can I recover my work after using the /clear command?

Yes, the skill includes a session catchup script that scans your project directory for existing planning files and git diffs to restore context in a new session.

File-Based Task Planning

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

byShawn-Guo-CN

0•

Productivity & Workflow

Organizes and tracks complex development tasks using persistent markdown files to ensure session recovery and consistent execution.

This skill implements a 'Manus-style' planning architecture that uses the filesystem as external working memory for Claude. By maintaining persistent files like task_plan.md, findings.md, and progress.md, it enables the agent to handle multi-step projects, complex research, and automated session recovery after context clears. It features a robust 3-strike error-handling protocol and strict '2-action rules' to prevent data loss during long-running tasks, making it essential for any high-complexity backend or frontend development workflow where state persistence is critical.

Key Features

01Automated creation of task_plan.md, findings.md, and progress.md for state management

02Persistent session recovery to resume work seamlessly after context resets or /clear commands

033-Strike Error Protocol for intelligent failure recovery and systematic debugging

040 GitHub stars

05Pre-tool hooks that automatically refresh the agent's context by reading the current plan

06Strict '2-Action Rule' ensuring research findings are saved to disk immediately

Use Cases

01Conducting deep research tasks requiring documentation of numerous web searches and file reads

02Managing multi-step software engineering projects that exceed standard context windows

03Maintaining continuity across multiple Claude Code sessions or development stages

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

Key Features

01Automated creation of task_plan.md, findings.md, and progress.md for state management

02Persistent session recovery to resume work seamlessly after context resets or /clear commands

033-Strike Error Protocol for intelligent failure recovery and systematic debugging

040 GitHub stars

05Pre-tool hooks that automatically refresh the agent's context by reading the current plan

06Strict '2-Action Rule' ensuring research findings are saved to disk immediately

Use Cases

01Conducting deep research tasks requiring documentation of numerous web searches and file reads

02Managing multi-step software engineering projects that exceed standard context windows

03Maintaining continuity across multiple Claude Code sessions or development stages

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