Overview

Supercharge your AI coding agents with planning-with-files, a robust, file-based planning system designed for persistent and crash-proof execution. Inspired by Meta's $2 billion acquisition, Manus, this skill ensures your AI's plans survive context loss, /clear commands, and unexpected crashes by storing critical planning documents (task_plan.md, findings.md, progress.md) directly on disk. It seamlessly integrates with over 60 AI agents via the standardized SKILL.md, offering deterministic completion gating and shared multi-agent state for complex, long-running tasks.

Key Features

• Persistent File-Based Planning: Critical planning documents (task_plan.md, findings.md, progress.md) are saved to disk, guaranteeing continuity across agent restarts, context resets, or crashes.
• Crash-Proof Execution: Plans are designed to withstand interruptions, ensuring that no work is lost and agents can resume seamlessly.
• Context Loss Resilience: Agents retain their planning state even after the context window is cleared or reset.
• Deterministic Completion Gate (v3+): An opt-in mechanism that ensures an agent's task is truly completed before proceeding, preventing premature task termination.
• Multi-Agent Shared State: Facilitates collaboration and synchronized progress across multiple AI agents by sharing state information on disk.
• Broad Agent Compatibility: Works with a wide array of AI coding agents, including Claude Code, Codex CLI, Cursor, Kiro, OpenCode, and many others, through the SKILL.md standard.
• Manus-Style Workflow: Emulates the proven planning methodology of the successful AI company Manus.
• Autonomous & Gated Modes (v3+): Introduces advanced options for managing long-running agent sessions with enhanced control and reliability.

Typical Use Cases

• Long-Running AI Development Projects: Ideal for complex coding tasks, refactoring efforts, or large-scale codebase analysis that span extended periods.
• Automated Code Generation & Maintenance: Ensures that agents can continuously work on code generation, bug fixing, or feature implementation without interruption.
• Multi-Agent System Orchestration: Enables sophisticated coordination between multiple AI agents for tasks requiring distributed problem-solving or collaborative development.
• Research & Experimentation: Provides a stable foundation for AI agents exploring complex algorithms or performing extensive data analysis where preserving intermediate states is crucial.
• Agentic Workflows with High Interruption Risk: Essential for scenarios where the AI agent might be taken offline or its context reset unexpectedly.