QIS-E-408 QISAGENT.COM THE INTELLIGENCE LAYER REV 2026-08-28 · BUILD 44
The Agent Stack · Constraints

Guardrails AI

Programmable validation and correction applied to model inputs and outputs.

QIS-E-408·Listed — not yet reviewed

The governance lens

Constraints — the category

Programmable rules applied to what a system may accept or emit, enforced in code rather than stated in policy, and capable of halting execution. Definition and why it matters →

Kill criteria expressed as code. A constraint that halts execution is enforceable in a way a policy document is not.

Why this framing

There is no shortage of writing about Guardrails AI. This register asks the question the general coverage does not: what does it contribute to governing an autonomous system that touches capital, and where does that contribution stop?

Status and disclosure

Listed — not yet reviewed. Structured facts and a governance lens, with no score attached. A full profile requires hands-on use, and we do not score what we have not run.

Disclosure

No commercial relationship. Guardrails AI is listed on editorial merit and has no input into this description.