How does session recovery work?

The skill includes a session-catchup script that analyzes existing planning files and git diffs to restore Claude's understanding of the project state after a context clear.

When should I use this skill?

It is recommended for any task requiring more than five tool calls, research-heavy projects, or any multi-phase development work where context retention is critical.

Does it automatically update my files?

The skill uses hooks to remind the AI to update files, but the core behavior is that the AI agent proactively writes to these files after significant actions or discoveries.

Where are the planning files stored?

While the skill templates reside in the plugin directory, your project-specific task_plan.md, findings.md, and progress.md files are stored directly in your project's root folder.

What is 'Manus-style' planning?

It is a workflow pattern that uses persistent files as 'external RAM' for the AI, keeping track of goals and discoveries to prevent context window saturation during complex tasks.

Planning with Files (Manus-Style)

Name: Planning with Files (Manus-Style)
Author: OthmanAdi

byOthmanAdi

•

11,059

生産性とワークフロー

Implements a persistent, file-based "working memory" system to manage complex multi-step tasks and research projects within Claude Code.

概要

This skill transforms how Claude handles complex tasks by implementing the Manus-style planning workflow, which uses persistent Markdown files as a form of external disk memory. By maintaining a structured task_plan.md, findings.md, and progress.md in the project root, the skill prevents context loss during long sessions, tracks errors meticulously, and allows for seamless session recovery after clearing context. It is essential for any project involving more than five tool calls, ensuring that goals, research, and progress remain accessible even when the AI's internal context window becomes saturated.

主な機能

11,059 GitHub stars
Three-strike error protocol to prevent repetitive failures and agentic loops
Automatic session recovery and context catchup after using /clear or restarting
Automatic hooks that display the current plan before critical tool executions
Persistent 'Working Memory' on disk using Markdown files for long-running tasks
Structured file templates for task planning, research findings, and session logging

ユースケース

Debugging intricate issues that require tracking multiple attempts and error logs
Conducting deep research tasks where findings must be preserved across sessions
Managing complex, multi-step software engineering projects with 10+ tool calls

概要

主な機能

11,059 GitHub stars
Three-strike error protocol to prevent repetitive failures and agentic loops
Automatic session recovery and context catchup after using /clear or restarting
Automatic hooks that display the current plan before critical tool executions
Persistent 'Working Memory' on disk using Markdown files for long-running tasks
Structured file templates for task planning, research findings, and session logging

ユースケース

Debugging intricate issues that require tracking multiple attempts and error logs
Conducting deep research tasks where findings must be preserved across sessions
Managing complex, multi-step software engineering projects with 10+ tool calls