Six tool calls, one boundary - an interactive replay of the financial-services example #25
zohebk8s
started this conversation in
Show and tell
Replies: 1 comment
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
I wanted to see the credit-risk example as a sequence rather than as terminal output, so I built a browser replay of it. Single HTML file, attached - no build, no server, works offline.
What you can do with it:
mnpitool denied before the assessment even startsThe one I find most worth showing people is the sanctions scenario. Step 2 is allowed - screening has to run - and it comes back with a UBO match, so CDD never clears. Steps 3 through 5 carry on and produce a rating, an exposure position, an IFRS 9 stage. Then step 6 is denied, on the state those steps produced rather than on anything in the request. The agent still attempts the write. It doesn't matter. That's the part that's hard to convey in a log.
What's real and what isn't:
Worth being precise about, since a simulator that quietly overstates itself is worse than no simulator:
Genuinely computed in your browser: the LEI and IBAN check digits (ISO 7064 MOD 97-10), and the SHA-256 audit chain. The tamper test breaks a real hash chain and recomputes it forward from genesis - it isn't a scripted animation, which is why one edit invalidates that entry and every entry after it.
Not real: the tool responses are the repo's fixtures. The Trust Record signature is
simulated, not hardware-rooted, and attestation runs in dev-mode. The page says so inline in all three places rather than in a footnote.Scenarios, thresholds, and regulation citations follow the
financial-servicesSharing in case it's useful to anyone explaining this to a non-engineering audience - that's what I built it for.
Happy to hear if I've got any of the policy semantics wrong.
All reactions