The difference
Prompt guardrails tell the model what it should or should not do. Acira evaluates the browser action the model is trying to perform and enforces policy before execution.
Prompt guardrails influence model behavior. Acira controls the browser action channel. Teams often need both, but they solve different security problems.
Use this page when deciding whether prompt guardrails are enough for browser-using agents.
Prompt guardrails tell the model what it should or should not do. Acira evaluates the browser action the model is trying to perform and enforces policy before execution.
Prompt guardrails are useful for instruction quality, refusal behavior, and high-level policy reminders. They are not a deterministic control over browser commands.
Acira helps when the agent is about to click, type, navigate, upload, download, run a script, or use a secret. Those are actions where a separate enforcement layer matters.
No. Acira complements prompt guardrails by enforcing policy on browser actions after the model proposes them.
Because untrusted pages and tool outputs can influence the model. Action-level control checks the command itself before anything changes in the browser.