For anyone whose AI assistant has let them down
You have seen it: the confident answer that turns out to be wrong, the "done" that was never checked, the detail that quietly changes between one reply and the next. AIQT is one skill you add to your assistant that stops all three. With it on, your assistant checks its work before calling anything done, says plainly when it does not know, asks when a decision is yours, and never changes anything quietly. You get answers you can rely on, and a clear record of why.
The one priority ordering, decided in advance. The plain-terms version is below, for the reader who wants it.
Where to start
One standard, three paths. Start where you are.
One copy and paste into Claude, ChatGPT, or Copilot. From then on, every conversation is held to the standard.
Install AIQTInstall the governance core into your project. It sets itself up, generates the file each coding agent expects, and writes only inside its own directory.
For developersOne shared standard across every assistant your team uses, with central policy and reporting when an organization needs it.
For teamsWhat changes
Think of the last time your assistant let you down. It was almost certainly one of three failures: a confident claim with nothing behind it, unchecked work called done, or a detail that changed between one reply and the next. AIQT holds your assistant to a standard on all three.
Every factual claim points back to a source, and when a guardrail catches something, your assistant tells you which guardrail and what it caught. You can see the reasoning, so you can check it.
When your assistant does not know something, it says so plainly. When a decision needs your judgement, it pauses and asks. A plausible-sounding invention never fills the gap.
AIQT requires that "done" rests on a check that actually ran. Your assistant verifies before it declares, so every "done" comes with a check you can point to.
No quietly weakened checks, no half-built results dressed up as finished, no invented names or citations. When something fails, you hear about it before anyone else does.
The standard
Four things you can expect from an assistant that follows AIQT. Together they are the top tier of the ordering, and they always come first.
What it says matches what is true. Every claim points to its source, every statement about the state of something rests on an observation, and "done" means a check actually ran. An unknown is stated as an unknown.
The work is what it appears to be. Nothing stubbed or mocked is dressed up as finished, no check is quietly weakened, no name or citation is invented, and nothing changes silently. Failures are surfaced.
The work meets a standard of craft: correct against the requirements, consistent with the conventions, and complete across every surface a change touches.
Trust is warranted by the record and granted by you, one evidenced claim at a time. Every claim traces to evidence, every override is logged with a way to revert it, and failures are reported honestly.
Decided in advance. When correctness and craft pull against being fast or being cheap, correctness wins, every time. That call is made once, up front, so nobody has to re-argue it under a deadline. The four facets in the top tier are co-equal, with no ranking among them. Just below them sits Progress: when the answer is already clear, decide and act rather than deliberate. Progress never means less checking.
Open source
AIQT is open source under Creative Commons Attribution-ShareAlike 4.0. The guardrails work across tools, and any improvement contributed back stays open under the same licence, so a fix one team shares can be used by every other team. Contribution is opt-in.