claude-code - 💡(How to fix) Fix Feature request: global project registry / cross-project memory index

Summary

When working across multiple projects (each in a different directory), Claude Code's per-project memory is isolated. If I open Claude Code from a top-level directory or a different project, it has no way to know what other projects exist, where they live, or what skills/scripts are associated with them.

Problem

• Per-project memory lives in ~/.claude/projects/<slug>/memory/ and is only loaded when Claude Code is opened from that project's directory.
• ~/.claude/CLAUDE.md is the only always-loaded file, but it's not designed as a structured registry.
• There's no built-in way for Claude to answer "where are my projects?" without the user telling it every time.

Proposed Solution

A global project registry — a structured index (e.g. ~/.claude/projects.json or a section in ~/.claude/CLAUDE.md) that maps:

• Project name → working directory
• Memory slug → where per-project memory lives
• Key skills / slash commands associated with each project

This would allow Claude Code, when opened from any directory, to:

1. Know what projects exist and where they are
2. Load relevant cross-project context when needed
3. Answer "where is the skill for X?" or "which directory is project Y in?" without user re-explanation

Workaround (current)

I manually maintain a Project Map table in ~/.claude/CLAUDE.md. It works but requires manual updates and Claude can't programmatically navigate it.

Use Case

I run several automation pipelines (radio news bulletin, podcast publisher, weather scripts) across 4 separate directories. Every new Claude Code session in a different directory starts cold with no knowledge of the other projects. A global registry would let Claude help me navigate and manage the full system from any entry point.