Secrets
Sensitive values the agent uses at run time, referenced by placeholder and resolved across the hierarchy.
Secrets (sensitive data) let the web agent use confidential values (passwords, API keys, tokens) without writing them into a test in plain sight. Each secret has three parts:
- Placeholder: the name you reference in a test, e.g.
DEMO_PASSWORD. - Value: the real data the placeholder stands in for. It's filled in only while a test runs, and never appears in your stored execution records.
- Description: a note about what the secret is for.
In a step you reference the placeholder. The real value is filled in later, only at the moment it's needed.
The AI model works with placeholders, not values
The agent that drives the browser runs on an AI model, and it plans every action from your test plus screenshots of the page. Your secret values are never part of what the model is given. It works only with the placeholder name and your description. The real value is substituted down in the browser, at the instant a field is typed, not by the model. And any text that flows back to the model has its real values scrubbed back into placeholders.
One caveat: the agent reads the screen. A value typed into a masked field
(a password shown as ••••••••) stays hidden, but a value typed into a
visible field renders as plain text and can appear in the screenshots the
model sees. Secrets keep your values out of your test definitions, your prompts,
and your stored results. They don't repaint what's already visible on the page.
The hierarchy: effective secrets
You can define secrets at three levels: on a regression, a test suite, or a test case. When a case runs, Auretis merges all three into one set, the effective secrets, keyed by placeholder name.
If the same placeholder is defined at more than one level, the closest level wins:
Test case → overrides → test suite → overrides → regression. A secret on the test case beats the same placeholder on its suite, which beats the same placeholder on the regression.
So you can set a shared default once on the regression and override it only where a specific suite or case needs a different value:
| Placeholder | Regression | Test suite | Test case | Effective value |
|---|---|---|---|---|
DEMO_PASSWORD | pw-default | — | — | pw-default (from regression) |
API_TOKEN | tok-shared | tok-staging | — | tok-staging (suite overrides) |
CARD_NUMBER | — | — | 4242... | 4242... (case only) |
A test case's detail page shows its effective secrets, the resolved set across all three levels, so you can see which value wins for each placeholder before you run.