MCP tools document a .codmir.json projectId fallback that never resolves — every call fails with "projectId is required"

The Codmir MCP server's tool schemas advertise a working-directory fallback for projectId that does not exist in practice. Tools including get_project_context, list_todo, search_tickets and create_ticket all describe the parameter as: "The project ID (defaults to the project linked via .codmir.json)". get_project_context and list_todo additionally declare projectId as NOT required. In reality the MCP server process has its own working directory, unrelated to the directory the user/agent is operating in. It therefore never reads the caller's .codmir.json, and the documented default silently does not apply. Impact: every agent that trusts the documented schema issues a no-arg call, receives a hard error, and must be told the project ID out-of-band. For an autonomous/background agent with no human present, this is a dead stop rather than a degraded path — the agent has no way to discover the ID the schema promised it could omit.