Glossary · Identity & access

Credential injection

Substituting a secret’s value inside the browser at the moment a field is filled — after the planner has already decided what to type there.

What is credential injection?

Credential injection is the timing property that keeps a secret out of the model. The planner reasons over a token; the executor resolves that token to a value inside the browser session at fill time; and the value is redacted from the returned action path, the live frames and the logs. Nothing between the vault and the input field holds the plaintext.

Why it matters

Where a credential is resolved decides who can see it. Resolving at plan time puts it in a prompt, a trace and a log; resolving at fill time puts it in one keystroke sequence inside an isolated session.

Related terms

See it in context

Back to the full glossary

Put your agent to work. Keep the decision.

Give your agent a real browser, authorize the accounts it may use, and keep control of every step. Repeated runs compile into skills that replay without a model.