planning-with-files

What problem does it solve? Long multi-step tasks cause AI agents to lose track of goals after many tool calls, since context windows are volatile and limited. This Skill implements Manus-style file-based planning, using markdown files on disk as persistent working memory so goals, decisions, and errors survive context resets and /clear events. ## Core Features & Use Cases - Three-File Planning Pattern: Creates task_plan.md for phase tracking, findings.md for research and decisions, and progress.md for session logs in your project directory. - Automatic Session Recovery: The session-catchup.py script detects unsynced context from previous sessions and recommends how to restore state after /clear. - Completion Enforcement Hooks: PreToolUse, PostToolUse, and Stop hooks re-inject the plan into context and block stopping until all phases are marked complete. - Use Case: When starting a research project or feature build requiring more than 5 tool calls, the agent writes a phased plan first, logs every error, and re-reads the plan before major decisions to avoid the lost-in-the-middle effect. ## Quick Start Ask the agent to plan this complex task using planning files before writing any code.