What problem does it solve? Claude's per-project memory normally lives outside the repo at ~/.claude/projects/<slug>/memory/, so it is never versioned, never shared with the team, and lost when switching machines. This Skill relocates that memory into the repository at memories/repo/ and links it back with a directory junction, making project knowledge committable, reviewable, and automatically loaded every session. ## Core Features & Use Cases - Memory-in-repo setup: Creates memories/repo/ with a MEMORY.md index skeleton (Feedback / Reference / Project / User sections) and seeds a frontmatter template for new memory files. - Automatic migration: If a real memory folder already exists under ~/.claude, its files are migrated into the repo first (never overwriting newer files) before being replaced with a junction. - Idempotent junction linking: Computes the project slug the same way the harness does and creates a directory junction (no admin rights required, unlike symlinks); re-runs are safe no-ops. - Use Case: A team lead clones a project on a new machine, runs the setup script once, and every team member's Claude sessions instantly load the same curated project context committed in git. ## Quick Start Ask Claude to set up memory in this repo so project knowledge is version-controlled and shared with the team.