Adjudication layer
Live demonstration — synthetic data

Their receipt. Our verdict layer.

A receipt chain can be perfect and the conduct inside it can still be unauthorized. Below is a chain that fully verifies — every action signed, every link intact — carrying four divergences from the authority its principal actually granted. This page finds them, measures how long each stayed open, totals the exposure that accumulated, and apportions a hypothetical loss across the overlapping failures. Everything runs in your browser.

All data on this page is synthetic. This is a demonstration, not a service. Meridian Example Corp, agent EX-7, the vendors and every amount are invented.

1The mandate

The authority document. It lives in a different system from the receipts — and that separation is the whole point. A receipt system records what it did; it does not hold the document that says what it was allowed to do.

held by: —

Mandate hash

2The receipts

A third-party receipt chain from System B, the platform the agent actually ran on. Ingested here as an input format, unmodified.

Timeline

#WhenActionDetailAgainst the mandate
Show the raw receipts (JSON, as received)
loading…
Format attribution. These samples are in the Agent Receipt Protocol format (reference implementation Obsigna) — a third-party open protocol. PlenaProof consumes it as an input. No endorsement or affiliation is implied. Validated against the upstream published schema; see the manifest in panel 4 for the version pinned here.

3Adjudication

Each receipt is evaluated against the mandate — never against the authorization the receipt asserts about itself. Every figure below is a button: open it to see the receipts and the mandate clause that produced it.

Step log

Divergence intervals

When each divergence opened, and when — if ever — it closed.

Accumulated exposure

The unauthorized amount that built up inside each open interval.

DivergenceIntervalExposure

Loss apportionment

Control failureAttributableShare%

4Determinism manifest

Re-running this computation on these inputs is bit-identical. That is the evidentiary point: a number nobody can reproduce is an opinion.

Evaluation order

    5Why this layer

    Receipt systems establish what happened and that the record has not been altered since. That is real work, and it is hard: signing keys held outside the agent process, hash-chained logs, tamper-evident structure. A chain like the one above does its job completely.

    What it does not do — cannot do, because it does not hold the document — is compare the conduct it recorded against the authority the principal actually granted. Nothing in a receipt chain measures how long an unauthorized condition stayed open, totals the exposure that accumulated while it did, or separates overlapping control failures when a loss lands on all of them at once. The chain above records a delegation, and records the payments that delegation enabled, and is entirely correct about both. It has no way to know the mandate never permitted the delegation.

    PlenaProof consumes any conformant receipt format and computes exactly that difference. The inputs are someone else's records. The output is a deterministic computation over an independently held mandate — not a legal determination, not a finding of liability, and not a substitute for the judgement of the people who have to decide what to do about it.

    What this page does and does not show. It records a comparison between a synthetic mandate and a synthetic receipt chain, and reports the arithmetic. It does not establish that any event occurred, that any party is at fault, or that any share is owed. The apportionment is the output of a stated rule applied to invented numbers. PlenaProof is not a law firm, a government agency, a bank, a licensed medical provider, or an official institutional partner.
    All data on this page is synthetic. This is a demonstration, not a service.