This Skill Gives OpenCode a Manus Brain

Join My Skool → https://clickmoney.io/gzkqjw9n-fajdlk Stop losing progress with your AI coding agents. Learn how to give OpenCode a Manus brain using a new GitHub skill for long term memory and planning. In this video, I break down why your current AI agent workflow might be failing and how to fix it with a dedicated memory layer. Why use Planning with Files for OpenCode? Vanilla OpenCode stores state inside the chat context. This leads to context stuffing, goal drift, and forgotten tasks. By using the Manus style planning skill, your AI agent treats the LLM context like RAM and your file system like a hard drive. This keeps the agent focused on long term goals even if you have to reset the session or hit token limits. Core Benefits Persistent State: Research and progress are stored in Markdown files rather than getting buried in chat history. Higher Success Rates: Benchmarks show task completion jumping from 6 percent to over 96 percent when using this planning loop. Easy Resumes: Start a new session and have the agent pick up exactly where it left off by reading the task plan. Reduced Context Bloat: Keep your context window slim by offloading research data to external files. Check it out here: https://github.com/othmanadi/planning-with-files Timestamps 00:00 Introduction to the OpenCode Planning Skill 00:13 Architecture differences between Vanilla OpenCode and Planning Agents 01:32 How the skill uses the file system like a computer disk 02:22 Step by step installation and project setup 03:31 Explaining the Task Plan, Progress, and Findings files 04:30 Demo 1: Complex research on AI fashion products 06:01 How to manually invoke the skill in the terminal 07:11 How to resume a task in a new session to save tokens 08:05 Demo 2: Compiling YC company data into structured tables 08:31 The internal agent loop: Read, Work, Write, and Reread 09:29 Performance Benchmarks: 6 percent vs 96 percent pass rate 10:41 Final takeaways for technical refactors and migrations