When is it best to use the Planning with Files skill?

It is recommended for any task requiring more than 5 tool calls, complex multi-step research, or projects where tracking errors and incremental progress is critical.

How does the 3-strike error protocol improve reliability?

It forces a logical progression: first diagnose the fix, then try a different tool or method, and finally rethink the entire approach, preventing the AI from repeating the same failing action.

Why use files instead of just relying on the context window?

The context window acts like volatile RAM which can become cluttered or lose detail; writing to disk provides 'permanent memory' that Claude can re-read to maintain perfect state during long sessions.

What are the three core files created by this skill?

The skill manages task_plan.md for phases and goals, findings.md for research and data storage, and progress.md for session logging and test results.

Planning with Files

Name: Planning with Files
Author: jukasdrj

byjukasdrj

•

生産性とワークフロー

Implements persistent, file-based task planning and research tracking to handle complex multi-step AI workflows.

概要

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.

主な機能

Automated plan verification hooks that check progress before critical tool executions
Persistent working memory using standardized markdown files (task_plan.md, findings.md, progress.md)
2-Action rule for immediate persistence of research and multimodal findings
1 GitHub stars
Manus-style planning logic optimized for complex, multi-step AI tool chains
Standardized 3-strike error protocol to ensure intelligent troubleshooting and escalation

ユースケース

Executing large-scale software migrations or complex architectural refactors
Conducting deep research projects requiring multiple web searches and data synthesis
Managing development tasks that span dozens of tool calls and require state recovery

概要

主な機能

Automated plan verification hooks that check progress before critical tool executions
Persistent working memory using standardized markdown files (task_plan.md, findings.md, progress.md)
2-Action rule for immediate persistence of research and multimodal findings
1 GitHub stars
Manus-style planning logic optimized for complex, multi-step AI tool chains
Standardized 3-strike error protocol to ensure intelligent troubleshooting and escalation

ユースケース

Executing large-scale software migrations or complex architectural refactors
Conducting deep research projects requiring multiple web searches and data synthesis
Managing development tasks that span dozens of tool calls and require state recovery