claude-config/memory/project_plugin_sync.md

name, description, metadata

name	description	metadata
project-plugin-sync	Claude Code plugins must stay synced across all servers via the claude-config repo (settings.json enabledPlugins)	node_type type originSessionId memory project e24d82e5-1bd0-48dc-99e6-9a4a434f0d2d

All of the user's Claude Code installations (multiple servers) must run the same set of plugins. The user explicitly wants newly installed plugins propagated to every machine.

Currently enabled plugins (in ~/.claude/settings.json → enabledPlugins):

• From claude-plugins-official (anthropics/claude-plugins-official, auto-installs on every machine): carta-investors, superpowers, github, claude-md-management, aws-agents, feature-dev, huggingface-skills
• Pre-existing: codex@openai-codex, warp@claude-code-warp (warp marketplace is in extraKnownMarketplaces).

Why: Plugins are declared in settings.json (enabledPlugins + extraKnownMarketplaces), which the user-github claude-config repo already tracks. The installed plugin cache under ~/.claude/plugins/ is NOT synced and should not be — each server re-fetches plugins from its marketplace based on the declaration.

How to apply: After installing/removing any plugin, run ~/Git/claude-config/sync.sh to push the updated settings.json to GitHub. On another server, git pull && ./install.sh, then start Claude Code — it auto-installs the enabled plugins from claude-plugins-official. Gap to watch: any plugin from a non-official marketplace (e.g. openai-codex) must have that marketplace added to extraKnownMarketplaces in settings.json, or it won't install on other servers (codex@openai-codex currently lacks this).