If you're working on anything that takes more than a handful of steps, this gives Claude persistent memory through three markdown files: task_plan.md for your roadmap, findings.md for discoveries, and progress.md for tracking what's done. The standout feature is session recovery. Run a Python script after you `/clear` the conversation and Claude picks up right where it left off by reading those files back in. It enforces a workflow: plan first, update after each phase, log errors, and never retry the same failed action twice. There's a three-strike protocol for debugging and templates to keep things structured. Basically turns your project directory into Claude's external brain so context survives across sessions.
npx -y skills add othmanadi/planning-with-files --skill pi-planning-with-files --agent claude-codeInstalls into .claude/skills of the current project.
Select a file.
mattpocock/skills