When should I use the planning-with-files skill?

This skill is designed for complex tasks involving more than 5 tool calls, deep research projects, or multi-phase builds where state management is critical.

How does this help with context window limits?

It offloads important data from the volatile context window to the local filesystem, allowing Claude to reference long-term goals and findings without exhausting its immediate memory.

What is the Three-Strike Error Protocol?

It is a systematic debugging approach: Attempt 1 is diagnosis and fix, Attempt 2 is a completely different method, and Attempt 3 is a broader rethink before escalating to the user.

What specific files does this skill manage?

It manages task_plan.md for phase tracking, findings.md for research storage, and progress.md for session logging and test results.

Manus-Style Agentic Planning

Name: Manus-Style Agentic Planning
Author: iseepatterns-jz

byiseepatterns-jz

生産性とワークフロー

Implements a file-based memory system for Claude to manage complex, multi-step agentic tasks without losing context.

概要

This skill equips Claude with a structured methodology for handling sophisticated projects by utilizing the filesystem as a persistent working memory. By mandating the creation of dedicated plan, findings, and progress files, it ensures that critical insights, research discoveries, and error logs are never lost to context window volatility. Inspired by high-performing agents like Manus, this pattern is essential for research-heavy tasks, large-scale refactors, or any operation requiring extensive tool usage and iterative decision-making.

主な機能

Structured phase tracking for complex project management
Periodic findings capture to prevent multimodal information loss
Persistent file-based memory (task_plan.md, findings.md, progress.md)
Automatic error logging and Three-Strike recovery protocol
0 GitHub stars
Mandatory 'Read Before Decide' workflow for goal alignment

ユースケース

Long-running debugging sessions requiring systematic state tracking
Deep research and information gathering across multiple web sources
Complex multi-step software architectural refactors

概要

主な機能

Structured phase tracking for complex project management
Periodic findings capture to prevent multimodal information loss
Persistent file-based memory (task_plan.md, findings.md, progress.md)
Automatic error logging and Three-Strike recovery protocol
0 GitHub stars
Mandatory 'Read Before Decide' workflow for goal alignment

ユースケース

Long-running debugging sessions requiring systematic state tracking
Deep research and information gathering across multiple web sources
Complex multi-step software architectural refactors