Why this question matters
A secret in a repository stays in the change history even after the file is deleted. Anyone with access to the repository then has access to the password, and a plant that installs the system on its own premises inherits that risk. The question to the vendor is therefore not about perfection, but about a mechanism that blocks the mistake before it reaches the history.
A good vendor answer has three parts: a policy, an automatic check and proof that the check has worked in practice or that an earlier mistake was removed.
Policy: where the values live
Our policy is that the code repository holds no environment variable files. Values are kept in CI/CD, in the container environment on the servers and in the default values of the developer's local stack. On the servers, the container environment supplies the whole configuration.
The list of required variables sits in the repository as a contract containing only the names, without values. An automated test in every pipeline checks that no required name is missing and that the file contains no equals sign.
During installation at a client site, the policy has one more dimension: secret values are created and stored on the plant's side, and the vendor neither knows them nor keeps them in its repository. Ask, therefore, who generates the installation passwords and keys, who rotates them and in which store they sit on the plant's server.
Automation: the scan in the pipeline
A secret scan in the pipeline blocks a password or key from entering the repository. This is what separates a mechanism from a declaration: a change containing a secret does not go further. Ask the vendor for the name of the tool, the scope of the scan, and whether it covers the history and not only the latest commit.
Ask also what happens after detection. A proper answer is that the change is blocked, the value is invalidated and an entry is made in the event register, not simply that the line is removed from the code.
Proof: what the vendor should show
A credible vendor shows the mechanism, not just a declaration: the pipeline configuration with the scan enabled, the scan result from the latest release and a description of the procedure after a secret is detected. It is also good when the vendor can point out that the list of required variables is a contract with names only, and that the scan blocks a repeat of the error.
Request three specific artefacts: an export of the pipeline configuration with the scan enabled, the scan report from the latest release, and the procedure for responding to a detected secret. The procedure should name the responsible person, the deadline for invalidating the value and how the client is notified if the secret concerned its installation. If any of these elements is missing, record it in the questionnaire as a gap to close before the contract is signed.
Proof can also be system behaviour: a missing variable ends with a message naming the missing setting, and a missing operating-mode variable stops the start instead of quietly enabling developer mode with a debugger.
Questions to paste into the questionnaire
The questions below can be pasted into the vendor questionnaire. Each has an answer that can be verified: by a document, a pipeline configuration export or a demonstration.
Add a deadline for the answer and a named person on the vendor's side. A general answer such as 'we apply good practices' is not proof.
- Does the repository contain files with environment variables?
- Where are the secret values kept on the servers and in the pipeline?
- Does the secret scan block a change in the pipeline, and does it cover the history?
- What happens to a secret after it is detected?
- Does missing configuration stop the start of the system?
Video
Frequently asked questions
Is a secret scan enough as the only safeguard?
No. It is one layer. It is complemented by a policy of keeping values outside the repository and by a test that enforces a list of names without values.
What distinguishes proof from a declaration?
Proof can be checked: the pipeline configuration, the scan result, the description of the procedure after a secret is detected. A declaration is an assurance without verification.
Does the absence of a secrets file mean there is no risk?
No. The values have to be stored somewhere, so ask in which store they sit and who has access to them.
How often should the question be repeated?
When choosing a vendor, before a major update, and after every security incident on the vendor's side.
Would you like to see our process?
We will explain how the secret scan and configuration without env files work, and answer your IT department's questionnaire.
Let's talk