What problem does it solve? Long agent sessions lose track of goals after many tool calls, and abstract plans drift away from the actual repository structure. This Skill keeps plans short, verifiable, and grounded in real files by persisting goals, decisions, and progress to Markdown files on disk. ## Core Features & Use Cases - File-anchored minimal planning: Produces plans of at most 3 steps tied to concrete paths, symbols, and the cheapest discovery tool (list_files, search_files, read_file). - Persistent planning files: Ships templates for task_plan.md, findings.md, and progress.md plus init scripts (Bash/PowerShell) to bootstrap a session. - Completion and catch-up automation: check-complete scripts verify all phases are done, and session-catchup.py detects unsynced context from previous sessions. - Use Case: Before editing code, ask the agent to plan the next change; it returns an entry point file, an initial tool, a 3-step plan, and a closure criterion instead of a long abstract strategy. ## Quick Start Ask the agent to plan the next concrete step for your task using real repository files and tools before making any edits.
