open-claude-designlisted

# Claude Design Use Claude Design as an external design workspace without loading its tool catalog into unrelated sessions. This skill owns access and synchronization; `open-claude-ui-design`, `open-claude-design-system`, and `open-claude-ui-review` continue to own product-design judgment and repository implementation. ## Use one transport Use the `open-claude-design` CLI from every coding agent, including Claude Code. This keeps discovery, path boundaries, etag checks, backups, capability redaction, and verification identical across hosts. Do not bypass it with a native Claude Design connector. Start every remote task with `open-claude-design status --json`. If authentication is missing or expired on a desktop host, tell the user that the first-use browser connection is opening, run `open-claude-design login`, then retry status after it succeeds. In CI, SSH, or a headless/dev-container runtime, do not run automatic or manual login inside the agent session: tell the user to run `open-claude-design login --manual` in their own interactive terminal, open its URL on a host browser, and paste the returned code into that terminal—not into the coding-agent chat. The browser flow is independent of the coding agent and API keys; never print, log, reconstruct, or ask the user to copy its tokens. On macOS the CLI stores its scoped credential in a dedicated Keychain item. On Linux and WSL2 it uses `~/.config/open-claude-design/credentials.json`, rejects symlinked paths, and require