Connecting a source system

Sign-off: approved by [Kyle] on [date]; recorded in approvals.md.

Use this for every system SCOPE connects to, such as Lytx, a routing or billing system, payroll or an HR roster. Copy the list into the integration's pull request and tick it there. The merged PR is the evidence. Finish it before the integration carries client data.

Who moves the data?

• The client, or we fetch with the client's login: add a line to the client's responsibilities in the system description.
• A third party (Fivetran, Zapier, a vendor's cloud): it is a subprocessor. Add it to the subprocessor list, collect its SOC 2 report yearly, and cover it in the DPA.

Checklist

• The data flow is written in the system inventory and the system description
• New fields are classified and have a retention period (data classification); take only what a screen needs
• Its key or password is in the vault and the secrets inventory, with an owner and a yearly rotation; never in git
• It writes through its own database role, limited to its own tables (as SCIM does), never the service-role key
• It is encrypted in transit and refused while the client's switch is off
• Every run is logged, and a failed or late run reaches a person
• It went through a normal pull request with a verify script
• The risk register was reviewed for it
• We know how to switch it off and revoke its key if the client or vendor leaves