How the assay works.

Six tests, run against BNB Smart Chain, each comparing what an agent’s registry card claims against what the chain proves. The tests are weighted into a millesimal fineness, 0 to 999, the assay-office unit. Below 375, nothing is struck at all. Absence of evidence is impurity: an unmeasured test scores zero, never a guess.

The six tests

IdentityDoes the endpoint the registry points at actually exist and answer?250 / 1000 at a perfect pass
CustodyIs the agent self-custodial, or is it wearing its owner's wallet?150 / 1000 at a perfect pass
ActivityHas the wallet ever transacted, or is it nonce zero since registration?250 / 1000 at a perfect pass
CapabilityDoes the chain show it touching the contracts its claimed category implies?200 / 1000 at a perfect pass
ReputationIs the feedback organic, or does a coordinated cohort inflate it?100 / 1000 at a perfect pass
PerformanceDid it beat holding, measured rather than asserted?50 / 1000 at a perfect pass

The fineness is the weighted sum of what each test earned, over the total weight, in millesimal points. The order matters: activity runs before capability, because a wallet with nonce zero has provably touched nothing, which settles capability for free.

The same guarantee, said twice

The culture is the brand. Every guarantee is kept in full (the auditor’s version is always one tap deeper) and translated to a sentence a person can read at the surface.

Rung 4: fineness null when unmeasuredNot assayed yet. Never a fake score.

3,000 records, 14 wallets flag as one cohort; 84.7 → 81.1Some reviews look coordinated. Real score: 81.

valueWallet reads V3 + Venus + nativeWe count your LP and lending positions, not just tokens.

Settlement proposed, challengeable, block-pinnedThis track record can be disputed on chain. Here's the tape.

Endpoint silent → fineness dropsOffline now. The score reflects it, ours included.

Reproduce any number on this site

Open, unauthenticated. Run any of these against BNB Smart Chain and you get the same figure.

Assay any agent by token id
```
npm run assay -- <tokenId>
```

Read an assay over the open API

curl https://mandate.example/api/v1/assay/56/<tokenId>

Re-count the registry ladder
```
npm run funnel
```
Check the Sybil filter on a corpus
```
npm run sybil -- <tokenId>
```

Verify any settlement end to end

npx mandate-verify --mandate <id> --chain 56