A worked architecture
Euro digital bond, five-year senior unsecured
Illustrative exampleSynthetic, no institutionNot an endorsement
A five-year senior unsecured bond issued directly onto a permissioned ledger operated by a market infrastructure. The ledger entry is the legally recognised record of ownership: there is no parallel register that could disagree with it. Coupons are calculated off-ledger by the paying agent and settled in central-bank money against the same ledger. Nothing crosses to another network, and no contract is upgradeable.
Everything below was produced by the same engine and rendered by the same components a member sees. Nothing here is a nicer version of the product.
The workflow
Eight steps, and the screen each one actually presents.
What follows is the member workspace, excerpted. The frames are the real controls in their real styling with the interaction removed, carrying this bond’s own answers, so what is shown here cannot drift away from what is built.
Step 1 of 8
The legal claim
What the holder actually owns, under which law, and where the token sits in relation to it.
Almost every expensive mistake in this market starts here. A token can be the legally recognised record, a mirror of one held elsewhere, or a receipt for a claim against somebody. Those three are not variations on a theme; they place ownership in different places and fail in different ways.
Step 1 of 8, The legal claimFrom the member workspace The first question, and the one every other answer depends on.What is the relationship between the token and the legal claim?
Until this is explicit, nothing downstream can be checked. Whether a transfer moves ownership, whether the register or the ledger wins in a dispute, and what a holder actually has if the platform stops, all follow from this one answer.
Your answerThe token is the legal assetThe ledger entry is itself the legally recognised record of ownership. There is no separate register that could disagree with it.Nothing is pre-selected. A default here would be a guess about somebody else's legal structure, and it would be wrong more often than it was right.
Step 2 of 8
Who controls what
For each thing the design must agree about, which record is authoritative when two disagree.
Systems do not disagree in theory, they disagree on a Tuesday afternoon. If two records are both authoritative and nothing ranks them, there is no answer to which one is right, and the reconciliation break has no owner.
Step 2 of 8, Who controls whatFrom the member workspace Ten things a design has to agree about. One row, and the hierarchy the answers build.How much exists
Which record states the total issued, across every network it lives on? Not the amount itself, the system that is right about it.
Which system is authoritativeThe ledger itselfNot the value. The system that is right about it.If two rows conflict, which winsThirdPosition is relative to the other rows, not a score.The hierarchy you have described
- 1instrument termsA global note or physical document
- 2holder positionThe ledger itself
- 3token supplyThe ledger itself
- 4cashThe calculation or paying agent
- 5corporate actionThe calculation or paying agent
Naming a system also gives it a place in the order. The two used to be separate controls, and a matrix could look complete while storing nothing at all.
Step 3 of 8
Who does what
The parties that operate the instrument, what they are in law, and what any of them can do alone.
An instrument is operated by institutions and people, not by a diagram. Naming them, and naming what each can do without a second pair of eyes, is how a control gap becomes visible before somebody finds it the hard way.
Step 3 of 8, Who does whatFrom the member workspace One of the 3 parties operating this bond, opened for editing.Who are theyPaying agentThe party, as your documents name them.What are they in lawAgent of the issuer for calculation and paymentTheir legal capacity, not their job title.How do they hold or reach positionsOperates the cash account against which coupons settleDirectly registered, through an omnibus account, by key control, or not at all.What approval do their actions needFour eyes on every payment instructionThe thing that stops one person acting alone. If the answer is nothing, say so.Rows open one at a time. A party has four fields, and five open rows is the wall of inputs this workspace exists to avoid.
Step 4 of 8
The standards stack
Which standards this architecture composes, and which job each one is doing.
Standards fail by being asked to carry something outside their scope. A token interface is not a bond taxonomy; a message is not a settlement. Choosing deliberately is how you find that out now rather than in production.
Step 4 of 8, The standards stackFrom the member workspace The catalogue proposes what this product usually needs. Every entry leads with what it does not do.Bond Data Taxonomy
RemoveICMA · ICMA-BDT-2
Machine-readable common language for conventional and DLT bonds across issuance, trading, settlement and distribution
Does not: Legal issuance documents prevail; BDT is not an executable lifecycle, token interface or settlement rail
ERC-7518: Dynamic Onchain Compliance for Security Tokens
Does not: Emerging model with large voucher, partition, payout and bridge surface
M3, draft or review
Selecting two alternatives from one group is allowed. It produces a finding rather than a refusal, because sometimes the honest answer is that the choice has not been made yet.
Step 5 of 8
Operating and unwinding
How the instrument is serviced over its life, who can change what, and how it ends.
The exit is the part nobody designs and everybody eventually needs. Detokenisation, migration, key loss and insolvency are ordinary operational events, and an architecture without answers for them is one bad day from an unrecoverable position.
Step 5 of 8, Operating and unwindingFrom the member workspace One of the 6 lifecycle steps, and the invariants the design rests on.Which stepissue, issuanceCreate legal/economic instrument and initial ledger representationWho performs itIssuerOne of the parties from the previous step.What starts itDrawdown under the programmeA date, an instruction, a threshold, or somebody deciding.How tightly is it controlledTwo people, separatelyTwo authorisations by different people. The usual bar for anything irreversible.What proves it happenedSigned drawdown notice and ledger entryWhat you would show an auditor or a supervisor afterwards.What must always be true
- The sum of ledger positions equals the issued nominal at all times.
- A transfer to an ineligible party cannot be recorded.
- No lifecycle state is overwritten without a superseding record.
Step names come from a library of 82, so two institutions describing the same operation use the same word. Choosing one shows what it is for.
Step 6 of 8
Settlement, network and evidence
What settles against what, on which network, and what you will be able to prove afterwards.
Finality is two separate things. A ledger can consider a transfer final while the law does not, and a delivery-versus-payment claim that has not distinguished them is a claim nobody has checked.
Step 6 of 8, Settlement, network and evidenceFrom the member workspace Answering one question can reveal several others, or keep them off the screen entirely.Does it cross networksNo, one networkAnswering yes adds seven decisions, each a real failure mode rather than paperwork.Personal data on the ledgerNo, nothing personal on the ledgerOnly references, commitments or credentials go on-chain.Yes, this is decided and written downLeave it unticked if it is still open. An honest gap is more useful than a tick nobody can stand behind.This bond settles free of payment, on one network, with no upgradeable contracts. So it is never asked the seven cross-network decisions, the seven on upgrade governance, or the three on finality. Different answers would add all of them.
Step 7 of 8
Risks and controls
What could go wrong, what stops it, and how far along that control actually is.
A risk nobody wrote down is a risk nobody owns. These carry into the passport, so whoever implements, audits or supervises the design is working from the same list you were, including the parts you have not solved.
Step 7 of 8, Risks and controlsFrom the member workspace One of the risks this design carries, with the control that answers it.What could go wrongThe ledger and the paying agent's records disagree about who was entitled at a record dateThe consequence, not the mechanism. What would somebody lose?What stops itRanked authority of record, with a daily reconciliation and a named owner for breaksThe control, and who operates it.How far along is that controlPlannedDesigned, not yet built.Which frameworks it answers toPFMI, DASCP-2024Optional. Separate several with commas.The list starts seeded from what no standard adequately covers for this product, always as unresolved. The workspace must not imply somebody has dealt with something it raised on their behalf. This row is one the architects added themselves.
Step 8 of 8
Review and approve
What the rules found, what is still open, and freezing a version somebody can defend.
An approved passport is a statement that this architecture was considered and signed off, pinned to the catalogue and rule set it was judged against. It is worth doing when the design is one you would defend, not when the form is merely full.
Step 8 of 8, Review and approveFrom the member workspace Freezing writes an immutable version, stamped with the catalogue and rule set it was judged against.Passport iddap_example_euro_digital_bondLowercase, prefixedVersion1.0.0Three numbers, separated by dotsConfidentialityConfidentialTravels with the document and every export of it.Freeze this version
A correction is a successor version. This one stays where it is, which is what makes its checksum worth anything.
And what is not here
This is an excerpt, not the instrument.
One question is shown from steps that hold several, one row from lists that hold many, and one architecture from a catalogue that supports thirteen product types. What sits behind the account is the working surface: the full question set as it adapts to your answers, the whole catalogue to compose against, the review workflow in which nobody approves their own work, and passports and exports carrying your organisation rather than a synthetic one.
The decisions
Eleven questions, and what this design answered.
Only what applies is asked. This bond lives on one network and has no upgradeable contracts, so it is never asked the seven cross-network decisions or the seven about upgrade governance. Answering differently would have raised them.
What is the relationship between the token and the legal claim?
Until this is explicit, nothing downstream can be checked. Whether a transfer moves ownership, whether the register or the ledger wins in a dispute, and what a holder actually has if the platform stops, all follow from this one answer.
The token is the legal asset
What kind of network does this run on?
It decides who can see what, who can halt it, and what is legally and operationally possible. A public network means anything written is visible to everyone, permanently.
A market-infrastructure platform
Is this design intended for production, or for a pilot?
It changes what is acceptable. A draft or pilot standard is a reasonable choice for an experiment and a recorded exception for production, because you are depending on something its own authors have not settled.
Production
Is the ISIN being used as the token's contract address, or as its on-chain identifier?
An ISIN identifies an instrument. A contract address identifies one deployment of one implementation. One instrument can have several token implementations and several technical identifiers, so treating them as the same thing breaks both registries the moment there is a second deployment.
No
Is the Digital Token Identifier being treated as the identifier of the legal instrument?
A DTI identifies a token's technical identity, not the instrument in law. Using it as the instrument identifier makes the legal asset indistinguishable from one particular technical representation of it.
No
How does this end, and what happens when something goes badly wrong?
Detokenisation, migration, insolvency, key loss and the platform being shut down are ordinary events over an instrument's life. An architecture with no answers for them is one bad day from a position nobody can recover.
Detokenisation returns holders to a book-entry position with the issuer's registrar. Migration to a successor platform is provided for in the terms, with identifier continuity through the instrument's own identifier. On issuer insolvency holders rank as senior unsecured creditors, unaffected by the ledger. Key loss is handled by the platform operator's recovery procedure under dual control. If the platform withdraws its service, the terms require ninety days' notice and a return to book-entry.
How do the asset and the cash change hands?
Delivery versus payment is a specific claim: that the two legs either both happen or neither does. Saying it without having established technical and legal finality, and what happens to a failed leg, is the most common overclaim in this market.
Free of payment
Which exact ISO 20022 messages, versions and usage profile are you conforming to?
There is no such thing as being ISO 20022 compliant. Conformance is always to a specific message, a specific version of it, and a community usage profile that says how the optional parts are used. Two systems that both claim ISO 20022 routinely cannot exchange anything.
seev.031.001.09 for corporate-action notification and semt.002.001.11 for holdings reporting, under the market usage guideline in force at issuance.
Will any personal data be written to the ledger itself?
On a public ledger this is permanent and visible to everyone, and it cannot be undone. Identity documents, borrower details and beneficial-owner records are the usual offenders.
No, nothing personal on the ledger
Does the instrument exist on, or move between, more than one network?
Crossing a network boundary means trusting something outside the ledger to carry a message. Supply can be duplicated, ordering can be replayed, and the destination may not enforce the same eligibility rules.
No, one network
Are you relying on a third-party audit as assurance?
An audit applies only to the exact code, configuration, scope and date it examined. Relying on one without recording those four means relying on a document that may no longer describe what you are running.
No
Who controls what
One record wins, and the order says which.
Two records both marked authoritative, with nothing ranking them, is the design defect that surfaces later as a reconciliation break nobody can adjudicate. Here the terms document governs what the instrument is, and the ledger governs who holds it.
| Order | For | The record that controls |
|---|---|---|
| 1 | instrument terms | A global note or physical document |
| 2 | holder position | The ledger itself |
| 3 | token supply | The ledger itself |
| 4 | cash | The calculation or paying agent |
| 5 | corporate action | The calculation or paying agent |
The stack
Six standards, each doing one job.
Bond Data Taxonomy
ICMA-BDT-2Machine-readable common language for conventional and DLT bonds across issuance, trading, settlement and distribution
Does not: Legal issuance documents prevail; BDT is not an executable lifecycle, token interface or settlement rail
Metamodel, methodology, repository and registered messages for payments, securities, funds, collateral and reporting
Does not: Not XML-only, not transport, not finality and not token compliance; registered messages are not themselves ISO standards and community profiles control
Identifies a financial or referential instrument
Does not: An ISIN is not a smart-contract address; one instrument can have multiple token implementations and DTIs
Legal Entity Identifier
ISO-17442-LEIGlobally unique legal-entity identity and reference data
Does not: LEI is not KYC, sanctions clearance, wallet ownership or authority for a particular action
Full ERC-20-compatible framework with identity registry, trusted issuers, claims, modular compliance, agents, freeze, forced transfer and recovery
Does not: Privileged roles, upgrades, identity data, registry availability and module composition dominate; Final is not regulatory approval
ERC-7092: Financial Bonds
ERC-7092Named bond terms, issuance, redemption and optional extensions
Does not: Not a complete BDT/CDM implementation, legal document or settlement system
ERC-1155 partitions, vouchers, locks, freeze, force transfer, recovery, payout and wrapping
Does not: Emerging model with large voucher, partition, payout and bridge surface
Still in draft, and this design is scoped to production
What the rules found
Reproducibility
catalogue 1.0.0 · rules 1.0.0 · f5bd64e4
1 finding from 20 rules. None of them blocks review. Every one is a deterministic rule result, not a model’s opinion, and the same answers always produce the same findings.
MonitoredHigh
Draft, pilot or legacy artifact cannot be a production default without exception, enhanced assurance and exit plan.
A draft, pilot or withdrawn artifact is a reasonable choice for an experiment. As a production default it means depending on something its own authors have not settled.
- You answered is this design intended for production, or for a pilot: production.
- ERC-7518 is M3
What would settle it: Record an exception with enhanced assurance and an exit plan, or choose a settled alternative.
Sources and rule reference
RULE-DRAFT-PROD
This example deliberately leaves that finding standing. One of the six standards is still a draft, and the design is scoped to production, so the engine raises it as something to monitor rather than something to fix: allowed, with a recorded exception, enhanced assurance and a documented way out. An example engineered to a clean sheet would teach that findings are something to avoid, when they are the reason the tool is useful.
The output
An Architecture Passport, pinned to what it was judged against.
With nothing blocking, this architecture can be frozen. The passport records the catalogue release and rule set it was approved under, so the same conclusions can be reconstructed years later rather than taken on trust.
- Standards profiled
- 7
- Open decisions carried
- 1
- Risks and controls
- 3
- Lifecycle functions
- 6
A member exports it as JSON, YAML, Markdown or CSV, each carrying the approval checksum so a recipient can verify they were sent that exact version. The passport also records that legal review has not started, because this is a synthetic example and a passport claiming legal review it has not had is the most dangerous thing the document could assert.