Readiness scan
The Responsible AI Readiness Framework has 38 checks across 7 dimensions. 32 of 38 of them can be judged from artefacts a repository actually contains — a risk register, contracts, guardrail configuration, a versioned adversarial set, IaC, drift thresholds, CI gates. This page runs those checks and shows you the evidence behind every point. No repository to hand? Answer the questions instead.
Your code never leaves this tab. The folder is read in your browser with the File API and scored here. No upload, no request, no storage, and no analytics on the result. A page that asks a company to be honest about its AI system has no business collecting the answer.
or choose it — nothing is transmitted either way.
Reads up to 30,000 files and 120 MB, skipping any single file over
400 KB; a larger tree is scored on what was read and says so.
No repo to hand? Score a or a .
A different promise, stated precisely. Here your browser downloads the files from GitHub, so GitHub sees the request — the repository and your IP address. Nothing still reaches this site: there is no server here to reach, and the scoring happens in this tab exactly as it does for a local folder. Public repositories only; no token is ever asked for.
github.com/owner/repo — or just owner/repo
A scan reads artefacts. It cannot tell you whether a guardrail blocks the attack it was written for, whether an alert reaches a person, or whether the escalation path has anyone at the end of it. Those are the checks marked human audit above.
Here is what that costs, measured rather than asserted. This page scores FintelliGuard — one of my own systems — at 95. The same framework, run against the same system by hand, in full, scores it 81: several dimensions come back amber once you ask whether the artefacts actually do their job. That fourteen-point gap is the audit.