Define how an operation can expose outputs

NEW PROPOSAL: frontend: type: tosca.nodes.WebApplication.WordPress properties: ... attributes: url: { get_operation_output: [ SELF, tosca.interfaces.node.lifecycle.Standard, create, url_from_script ] } interfaces: tosca.interfaces.node.lifecycle.Standard: create: implementation: scripts/frontend/create.sh # This script outputs two env vars: "url_from_script" and "data_dir_from_script" inputs: ip_address: { get_attribute: [my_server, ip_address] } configure: implementation: scripts/frontend/configure.sh inputs: data_dir_from_script: { get_operation_output: [ SELF, tosca.interfaces.node.lifecycle.Standard, create, data_dir_from_script ] } outputs: url_output: value: { get_attribute: [ frontend, url ] } PREVIOUS PROPOSAL: frontend: type: tosca.nodes.WebApplication.WordPress properties: ... interfaces: tosca.interfaces.node.lifecycle.Standard: create: scripts/frontend/create.sh inputs: ip_address: { get_attribute: [my_server, ip_address] } outputs: url_from_script: { set_attribute: [frontend, url] } Staying with the recommendation of environment variables, the "url_from_script" output is an env var that was exposed by the operation This example shows the concept but we might need to think of a better syntax because normally, i.e. in most or all cases, the "thing" before the colon gets assigned the "thing" after the colon.