Cedulon: An Audit Layer for Agent-to-Agent Commerce

Cedulon: An Audit Layer for Agent-to-Agent Commerce VERAX TEKNOLOJI LIMITED SIRKETI

Turkey e.dogru@cedulon.com

sec Cedulon agent receipt policy SCITT This document defines the Cedulon Protocol, an audit layer for agent-to-agent commerce. Payment rails such as HTTP 402 flows (x402) and mandate protocols (AP2) already move value. They do not, by themselves, produce a fail-closed policy check and a signed spend receipt that a verifier can reconcile against a rail extract. Cedulon specifies a Trade Manifest (signed offer before payment), a Policy Decision Point with default deny, a Spend Receipt (COSE/CWT claim set after a gated payment), epoch checkpoints, and rail-extract reconciliation. The reconciliation shows that no settlement on the extract lacks a receipt and no settled receipt is absent from the extract. That result is unconditional only when the verifier pins the rail key out of band and states the period under audit; otherwise the document requires it to be reported as conditional. Checkpoints carry the suppression guarantee, so the document profiles the checkpoint as a Signed Statement, gives the verification algorithm a step that consumes the transparency receipts returned for checkpoints, names what a witness holding a checkpoint the presented chain omits reports, brings equivocation within reach by comparing recorded copies against the presented chain, and states how checkpoint totals may be withheld without withholding the fact that they were. This revision states the trust roots the earlier ones left implicit: no signed object may be verified against a key it carries itself, and a presented Trade Manifest must be bound both to the receipts that name it and to the terms those receipts claim. It also names a threat no adversary causes, a settlement recorded on a rail with no receipt behind it. It also defines a Dispute Evidence Bundle (evidence, not an award) and optional SCITT anchoring. Cedulon is not a competitor to x402 or AP2; it sits above them.

Introduction Note to Readers: This document is submitted as Informational. The author's eventual intended track, if the work is taken up, is a Standards Track profile of COSE and CWT for agent-spend receipts. This -03 does not claim IETF consensus. Agents can now pay. Open HTTP 402 protocols attach stablecoin settlement to ordinary requests. Card networks and processors issue agent-scoped tokens. Google's Agent Payments Protocol (AP2) binds user intent to signed mandates. What is missing is an interoperable audit layer: a machine-checkable answer to "was this spend allowed by policy, against which offer, and what bytes were delivered?" Without that layer, a prompt-injected or looping agent can drain a rail that has already accepted a valid signature. A counterparty can ship the wrong artifact. A transparency log, if used at all, is proprietary. Cedulon fills that gap. It does not clear funds, hold custody, or operate a payment facilitator. An optional escrow actor is defined only as a third-party role interface (). Implementations of this specification MUST NOT take custody of funds or operate escrow (MUST-T8-custody). The control is an old one. Reconciling an internal ledger against an external statement is what double-entry bookkeeping made routine, and signing the artifacts on both sides is Grigg's triple-entry idea . Neither is claimed here. What this document contributes is an open wire profile for that control in a setting where the parties are software: a COSE receipt shape, an extract shape, a checkpoint chain, and a verification algorithm precise enough that two implementations reach the same finding on the same evidence. The novelty is interoperability, not the idea. Neighbor drafts are complementary, not substitutes. draft-bates-atp covers tamper-evident causal lineage as a signed DAG. Cedulon is the completeness layer: a spend that never produced a receipt is visible when an authenticated rail extract is reconciled (MUST-T10-1).

Terminology The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 when, and only when, they appear in all capitals, as shown here. The following terms are used:

Trade Manifest:: A signed statement produced before payment. It binds a description of goods or service, price, currency, acceptance-criteria hash, cancel condition, expiry, and an optional AP2 mandate reference.
Policy Decision Point (PDP):: The function that evaluates a structured spend request against stored policy. The default is deny.
Spend Receipt:: A signed statement produced after a gated payment attempt. It binds payer, payee, amount, currency, policy hash, manifestHash or an explicit noManifest flag, rail payment reference, timestampMs, nonce, prevReceiptHash, and outcome.
Receipt Issuer:: The party that signs Spend Receipts.
Anchor:: An optional SCITT Transparency Service that registers a signed statement and returns a COSE receipt .
Dispute Evidence Bundle:: A package of the Trade Manifest, the Spend Receipt, and a delivery hash. It is evidence for a later human or legal process. It is not an arbitral award and not an escrow release.
Decision Token:: A portable, single-use PDP allow encoded as COSE_Sign1. The claim set binds requestHash, policyHash, expiryMs, nonce, and singleUseId. See .
Rail Extract:: An authenticated list of settlement records for one account, one rail, and one time window. See .

Architecture Cedulon has three control-plane objects and one optional log: PDP --allow/deny--> x402/AP2 rail | v Receipt Issuer --> Spend Receipt | v Anchor / SCITT (optional) ]]> The payer agent never talks to the rail except through an adapter that calls the PDP first (MUST-T5-1).

Policy Decision Point The PDP evaluates structured fields only (MUST-T1-1): amount, currency, payee, tool identifier, nonce, optional manifest hash, and evaluation time. It applies limit, velocity, and scope checks (MUST-T2-1, MUST-T2-2). If the PDP is unreachable, uninitialized, or throws, the result is deny (MUST-T2-3). Denied attempts do not increment success counters (MUST-T2-4). An allow produces a Decision Token whose requestHash covers six fields: amount, currency, payee, tool, nonce, and manifestHash (MUST-T3-4, MUST-T6-1). The token is a COSE_Sign1 object (MUST-T6-4), is single-use (MUST-T6-2), and MAY be carried to the adapter that performs settlement.

Receipt Issuer After the adapter attempts settlement (success or a recorded deny that still needs an audit trail for an allowed-then-aborted path), the Receipt Issuer signs a Spend Receipt over a canonical encoding (MUST-T4-1). Verifiers reject bad signatures and byte mismatch (MUST-T4-2).

Anchor / SCITT Parties MAY register the signed receipt (or a privacy-preserving hash encoding) as a SCITT Signed Statement and attach the COSE receipt (MAY-T4-6). This document does not operate a Transparency Service.

Trade Manifest A Trade Manifest is the commerce analogue of a promise: it is issued before value moves. It is conceptually symmetric to a later Spend Receipt (promise then proof), and it MAY carry an AP2 mandate hash so that user intent and the Cedulon offer stay linked (SHOULD-T8-5). A Trade Manifest MUST bind all of the following (MUST-T8-1):

goods or service description
price (integer minor units, encoded as a decimal string matching 0|[1-9][0-9]*)
currency (ISO 4217 alphabetic or a documented token identifier)
acceptance-criteria hash (SHA-256 of the exact delivery bytes or of a declared schema instance)
cancel condition (opaque string agreed by the parties)
expiry (POSIX milliseconds, expiresAtMs)

It MAY include ap2MandateHash. The corresponding CBOR label is always present; a missing mandate is encoded as CBOR null. The manifest is COSE_Sign1 over a deterministic CBOR claim map (). manifestHash is the SHA-256 of the signed COSE bytes (MUST-T8-7). A spend bound to a manifest MUST be denied if the requested amount or currency differs from the manifest (MUST-T8-2) or if the manifest is expired (MUST-T3-3). A spend that is not bound to a verified manifest MUST be marked noManifest on the receipt and MUST still pass limit, velocity, and scope checks (MUST-T1-2). An implementation MAY refuse all noManifest spend (MAY-T1-4).

Spend Receipt The Spend Receipt claim set is carried in COSE_Sign1 wrapping a CWT-compatible map . New receipts MUST use the COSE profile (). Claims (MUST-T4-3, MUST-T4-4, MUST-T4-7):

Claim	Description
payer	Payer agent identifier
payee	Payee identifier
amount	Minor units as a decimal string `0\|[1-9][0-9]*`
currency	Currency identifier
policyHash	SHA-256 of the canonical policy document (lowercase hex)
manifestHash	SHA-256 of the signed manifest COSE bytes, or null when `noManifest` is true
noManifest	Boolean; MUST be true if and only if `manifestHash` is null
x402PaymentRef	Rail payment reference, or null
timestampMs	POSIX milliseconds
nonce	Unique spend nonce; at least 128 bits of randomness; unique in the issuer scope
prevReceiptHash	Previous receipt hash, or null for the first receipt (`SHOULD-T4-5`)
outcome	`settled` or `aborted`

A receipt with outcome = settled MUST have a non-null x402PaymentRef (MUST-T4-7). An aborted receipt MUST NOT be added into checkpoint totals. All twelve labels in are always present. An empty optional value is encoded as CBOR null, never by omitting the label. receiptHash is the SHA-256 of the receipt's signed COSE bytes, encoded as lowercase hex. Verifiers MUST reject a receipt if the signature fails or if the decoded claim map does not match the presented claims (MUST-T4-2).

Optional payee countersignature A payee MAY attach a countersignature over the issuer's signed Spend Receipt (MAY-T8-9). The profile uses a detached COSE_Sign1 whose payload is a CBOR map with a single private-use label:

Label	Claim	CBOR type
-70401	receiptCose	bstr (exact issuer COSE_Sign1 bytes)

The countersignature uses the header profile in and content type application/cedulon-countersign+cbor. This is a second Sign1 object, not RFC 9052 Countersignature0 (unprotected-header label 11). Countersignature0 would write into the issuer object and change receiptHash after issue, breaking the receipt chain. A detached Sign1 keeps the issuer bytes stable, reuses kid and content-type, and is absent by simply omitting the sibling object. Absence of a countersignature MUST NOT invalidate the issuer receipt (MAY-T8-9). If a countersignature is present, a verifier MUST reject it when the signature fails, when kid or content type does not match the configured payee key, or when label -70401 is not the issuer COSE bytes (MUST-T8-8). The identifier countersign-bad SHOULD be used for this condition. A Dispute Evidence Bundle that includes a verified countersignature has stronger evidence that the payee accepted those bytes; the bundle is still not an award (MUST-T8-4).

COSE Profile This profile uses deterministic CBOR Section 4.2.1 (definite lengths, shortest integer form, map keys sorted in bytewise lexicographic order of their encoded keys). Implementations MUST encode only the types used by Cedulon claim maps: null, bool, unsigned and negative integers, UTF-8 text, byte strings, arrays, and maps (MUST-T4-1).

Claim labels Registered CWT claims are not required in -02. Cedulon uses CWT private-use integer labels less than -65536 so that the profile does not occupy the 100-110 registry range. Receipt labels (MUST-T4-3, MUST-T4-4, MUST-T4-7):

Label	Claim	CBOR type
-70001	payer	tstr
-70002	payee	tstr
-70003	amount	tstr
-70004	currency	tstr
-70005	policyHash	tstr (lowercase hex SHA-256)
-70006	manifestHash	tstr / null
-70007	noManifest	bool
-70008	x402PaymentRef	tstr / null
-70009	timestampMs	uint
-70010	nonce	tstr
-70011	prevReceiptHash	tstr / null
-70012	outcome	tstr (`settled` / `aborted`)

Checkpoint labels (MUST-T11-1):

Label	Claim	CBOR type
-70101	epoch	uint
-70102	startMs	uint
-70103	endMs	uint
-70104	receiptCount	uint
-70105	chainHeadHash	tstr / null
-70106	totals	map tstr -> tstr / null
-70107	prevCheckpointHash	tstr / null

Manifest labels (MUST-T8-1):

Label	Claim	CBOR type
-70201	description	tstr
-70202	amount	tstr
-70203	currency	tstr
-70204	acceptanceCriteriaHash	tstr
-70205	cancelCondition	tstr
-70206	expiresAtMs	uint
-70207	ap2MandateHash	tstr / null

Decision Token labels (MUST-T6-4):

Label	Claim	CBOR type
-70301	requestHash	tstr
-70302	policyHash	tstr (lowercase hex SHA-256)
-70303	expiryMs	uint
-70304	nonce	tstr
-70305	singleUseId	tstr

COSE_Sign1 headers The protected header MUST be a deterministic CBOR map containing (MUST-T4-1, MUST-T4-8):

1 (alg) = -19 (Ed25519, ; the generic EdDSA value -8 from is deprecated for this profile)
3 (content type) = a tstr that distinguishes the payload: application/cedulon-receipt+cbor, application/cedulon-checkpoint+cbor, application/cedulon-manifest+cbor, application/cedulon-decision+cbor, or application/cedulon-countersign+cbor
4 (kid) = bstr, mandatory. The profile computes kid as the first eight bytes of SHA-256 over the issuer's SubjectPublicKeyInfo DER. A verifier MUST obtain the public key from an authenticated channel (preconfigured issuer set, directory, or transparency statement) and MUST reject a message whose kid does not match that key.

The unprotected header MUST be empty. The payload MUST be the CBOR encoding of the claim map. The signature is Ed25519 over the COSE Sig_structure ["Signature1", protected, h'', payload].

Decision Token A Decision Token is the portable encoding of a PDP allow. It is COSE_Sign1 with the header profile in and the labels in . All five labels are always present (MUST-T6-4). requestHash MUST be the six-field hash defined for the PDP (MUST-T6-1). policyHash MUST be the SHA-256 of the canonical policy document the PDP evaluated. expiryMs is a Unix time in milliseconds after which the token MUST be treated as expired (SHOULD-T6-3). nonce is the request nonce. singleUseId is the identifier consumed on the first settlement attempt (MUST-T6-2). A party that accepts a Decision Token MUST reject it if the signature fails, if kid does not match a configured PDP key, if the content type is not application/cedulon-decision+cbor, if the decoded claim map does not match the presented claims, or if expiryMs is in the past (MUST-T6-5).

Rail Extract Profile A verifier checks completeness against a rail extract, not against the issuer's own receipts alone (MUST-T10-7).

Record schema Each settlement record MUST contain:

Field	Type
ref	tstr (rail payment reference)
amount	tstr matching `0\|[1-9][0-9]*`
currency	tstr
timestampMs	uint

Scope An extract is scoped to one account identifier, one rail identifier, and one half-open time window [windowStartMs, windowEndMs).

Authentication The mock rail in the companion implementation signs the extract with Ed25519 over a canonical encoding of the scoped body. A verifier MUST obtain the extract from the rail or from a signature the rail published (MUST-T10-7). A deployment that cannot do so is running the reconciliation against evidence it did not obtain independently, and MUST report the guarantee as conditional. A signature on an extract proves internal consistency, not origin: a key generated by whoever produced the object verifies against itself. The verifier therefore MUST obtain the rail's public key out of band and MUST verify the extract signature against that key rather than against any key the extract carries (MUST-T10-8). A verifier that holds no such key MUST treat the guarantee as conditional. Keys are compared as bytes. A verifier MUST compare the pinned key and the key that signed the extract by their SubjectPublicKeyInfo DER encoding, not by any text encoding of it, so that the same key presented in a different envelope still compares equal (MUST-T10-9). A pinned key the verifier cannot decode is a fault in the verifier's own configuration, not evidence about the extract, and MUST be reported as trust-key-unreadable rather than as a key mismatch. What a verifier emits for an extract it cannot authenticate depends on whether it stated an expectation. With no pinned key the verifier has asserted nothing, so an extract that does not carry a verifiable signature is unauthenticated-extract, a warning: completeness findings may still be computed, but the guarantee is conditional on the extract being authentic. With a pinned key the verifier has asserted what it requires, and an extract that fails to meet it is a failure rather than a caveat; see the verification algorithm for which finding applies. -00 defined only the first case, and readers of -00 should note that this revision makes the pinned case fail closed. See .

Scope agreement An extract declares a window and carries settlement records. The two MUST agree: a verifier MUST report every settlement record whose timestampMs falls outside [windowStartMs, windowEndMs) as extract-scope-mismatch, identified by that record's ref (MUST-T10-10). This check is about the extract's internal consistency and MUST be performed whether or not a rail key is pinned. A verifier that knows which account, rail, and window it is auditing MUST also check the extract against that expectation and MUST fail closed when the extract does not cover it (MUST-T10-11). An extract for another account or rail, or one whose window does not span the period under audit, cannot support a completeness claim about that period. A verifier that states no period leaves the extract free to define one, and an extract that reports on a millisecond balances as easily as one that reports on a month. Pinning a key establishes who signed; only a stated period establishes what the signature had to cover. A verifier that has not stated the period under audit therefore MUST emit unstated-audit-window and MUST treat the guarantee as conditional (MUST-T10-15), whatever else verifies.

Trust roots states the rule for one object: a signature proves internal consistency, not origin, so the verifier obtains the rail key out of band and checks the extract against that key rather than against any key the extract carries (MUST-T10-8). -02 already required a verifier to obtain the public key from an authenticated channel and to reject a kid that does not match that key (MUST-T4-8). What -02 did not carry was the verification algorithm, the separate root inputs, and the error semantics that name a missing or mismatched pin. This section states those for every signed object in the profile. The gap is not theoretical. A verifier that checks a Spend Receipt against the key the receipt carries accepts a receipt signed by any key at all, including one an attacker minted for the occasion. Such a receipt matches a settlement the attacker was never authorised to make, the settlement stops looking uncovered, and the audit reports nothing. The completeness property in is then computed over evidence that answers to nobody.

The issuer root A verifier MUST obtain the issuer's public key out of band and MUST verify Spend Receipt and epoch checkpoint signatures against that key rather than against a key the object carries (MUST-T4-9). A verifier that holds no such key and is presented with any Spend Receipt or epoch checkpoint MUST treat the completeness guarantee as conditional and SHOULD report the condition; the identifier unauthenticated-issuer is used for it in this implementation. The condition names those two objects because an audit given no receipts and no checkpoints rests on the extract alone. There the absent issuer root withholds nothing, and warning about a root the audit never consulted would spend the warning where it carries no information. Reporting a mismatch is not sufficient on its own. A receipt that does not answer to the pinned issuer key MUST NOT count as coverage for the settlement it names, and the settlement MUST still be reported as uncovered (MUST-T4-10). A verifier that reports the mismatch and then lets the receipt match the settlement anyway has described the attack in its output while still concluding that the books balance. Keys are compared as bytes, by their SubjectPublicKeyInfo DER encoding, on the same terms as MUST-T10-9. A pinned issuer key the verifier cannot decode is a fault in its own configuration and MUST be reported as trust-key-unreadable rather than as a mismatch against the objects; where no pinned key can be decoded at all, nothing is attested and the verifier MUST NOT fall back to accepting the keys the objects carry (MUST-T4-11). Falling back is how a mistyped configuration becomes a bypass. A verifier MUST accept an issuer root that is a set of keys rather than a single key (MUST-T4-12). An issuer that rotates its key mid-window otherwise produces a finding against every honest receipt signed by the retired key, and the reachable way out of that is to stop pinning, which is the opposite of what the pin is for. The same acceptance applies to a publisher pin, a witness pin, and a rail pin: a verifier MUST accept each of those roots as a set of keys, so a rotation inside the window does not force it off the pin.

The payee root The optional countersignature in travels beside the issuer signature without being covered by it. Anyone holding an honest receipt can append a countersignature of their own, so a verifier that checks it against the key carried next to it learns only that some key signed something. A countersignature MUST NOT be treated as evidence that the payee approved the payment unless it verifies against a payee key the verifier obtained out of band (MUST-T4-13). Without such a key the verifier SHOULD report the condition and MUST treat the guarantee as conditional. Naming a payee key states an expectation, and an expectation that only fires when the evidence is present can be cancelled by deleting the evidence. Where a verifier has pinned a key for a payee, a settled receipt naming that payee and carrying no countersignature MUST be reported (MUST-T4-14). Otherwise an attacker removes their own failed forgery and the report returns to unconditional.

The witness root describes what a transparency witness adds. An inclusion receipt checked against the key it carries says that some log is internally consistent, and a log is cheap to invent. A verifier MUST obtain the transparency service's public key out of band and MUST verify inclusion receipts against it (MUST-T11-15). Inclusion receipts that cannot be checked that way MUST NOT be used as evidence, in either direction: they cannot establish that a checkpoint was anchored, and they cannot establish that one was withheld. A verifier SHOULD report that it left them out. A log holds statements from everyone who uses it. A statement held by a pinned log MUST additionally answer to the issuer root before it counts as something that issuer published (MUST-T11-16); otherwise another user's epoch, sitting in a shared log, reads as this issuer publishing two checkpoints for one epoch. Anchoring and withholding are different claims and need different evidence. Establishing that a checkpoint was logged needs only the statement hash. Establishing that an issuer withheld one needs to know whose statement it is, which an inclusion receipt carrying no statement body cannot say. Such a receipt MUST NOT be used to report a withheld checkpoint, and MUST NOT be silently discarded either: a real withholding must not be buried by removing the body (MUST-T11-17).

The decision root A Decision Token is issued by the policy decision point and consumed by the same deployment. The consumer therefore holds the key it signs with and has no reason to ask the token which key to check it against. A consumer MUST verify a Decision Token against its own issuing key and MUST NOT accept one it cannot check that way (MUST-T6-6).

The manifest root A Trade Manifest is optional. A deployment that presents none is not missing a root, and this requirement does not make such an audit conditional. The forbidden case is the other one: a manifest is presented, and the verifier accepts it because the key travelling inside it verifies against itself. A verifier that is presented with a Trade Manifest MUST obtain the publisher's public key out of band and MUST verify the manifest signature against that key, not against a key the manifest carries (MUST-T4-15). A verifier without such a key that is presented with a Trade Manifest MUST report the completeness guarantee as conditional and SHOULD report the condition; the identifier unauthenticated-manifest is used for it in this implementation. An audit presented with no Trade Manifest is not made conditional by this requirement. A pinned manifest key the verifier cannot decode is a fault in its own configuration and MUST be reported as trust-key-unreadable rather than as a mismatch, on the same terms as MUST-T4-11. A manifest that does not verify against a readable pin MUST be reported as manifest-key-mismatch and MUST fail the audit. Falling back to the key the manifest carries is how a presented document becomes a bypass. Attribution is one question and coverage is another, and a root that answers only the first leaves the document doing work it was never spent under. A verifier presented with a Trade Manifest MUST compare the manifest hash to the manifestHash of the receipts presented to the audit and MUST report a manifest that no presented receipt references (MUST-T4-17); the identifier manifest-covers-no-receipt is used for it in this implementation, and the completeness guarantee is conditional. The comparison runs against those presented receipts, including aborted ones, and is made before any extract window is applied; a hash on an aborted receipt or on a receipt outside the extract window still counts as a reference. This requirement asks whether any receipt names the terms, not whether a settlement in the window was made under them. A correctly attributed manifest travelling beside a set of receipts marked noManifest states terms nothing presented was spent under, and a report that stays silent about it reads as terms-backed when it is not. This requirement does not reach the audit presented with no Trade Manifest, which remains a deployment choice under MUST-T1-2. Naming a manifest is not obeying one, and that is the third place this document has had to say the same thing twice. A verifier presented with a Trade Manifest MUST compare the amount, the currency and the settlement time of every presented receipt that names it against the manifest's amount, currency and expiry, and MUST report a receipt that departs from them (MUST-T8-9); the identifier manifest-terms-mismatch is used for it in this implementation and the audit fails. Only receipts that name the manifest are measured against it: reading the terms onto a receipt that never claimed them would invent a violation the payer did not commit. The rules being enforced here are not new. A gate already refuses a bound spend whose amount or currency differs from the manifest (MUST-T8-2) and one made against an expired manifest (MUST-T3-3). Both were written for the point where money moves, and an audit reads the record after that point, where the gate is no longer present to be asked. Without a counterpart a receipt can carry the hash of terms it breaks and the report still calls the books balanced. This is a finding rather than a condition on the guarantee: a verifier that reports it held every root it needed, and the statement it is making is unconditional. The gate answers differently from the audit, and the difference is deliberate. A policy decision point presented with a Trade Manifest it cannot attribute MUST refuse the payment rather than settle and record the doubt (MUST-T4-16). An audit describes what it found and may say the result is conditional; a gate decides whether money moves, and a settled payment carrying the hash of terms nobody authorised cannot be withdrawn by reporting it afterwards. The receipt would record those terms as agreed.

What the roots do not cover A verifier that supplies none of these roots is not making an error, and this document does not require it to. It is making a weaker statement, and the guarantee it reports must say so. With no issuer key nothing distinguishes one submitted receipt from another, so conditions computed across the submitted set - two receipts claiming one settlement reference, for instance - cannot be attributed to anyone and are reported as conditions of the submission rather than as failures of a party.

Reconciliation and Epoch Checkpoints Completeness is the property that, given an authenticated rail extract, every settlement in the extract has a matching settled Spend Receipt, every settled receipt has a matching settlement, receipt and checkpoint hash chains verify, and checkpoint totals equal the sum of settled receipts in the checkpoint window. If a spend occurred without a receipt, the missing receipt is itself the evidence (MUST-T10-2). A checkpoint published with its totals withheld () cannot contribute the last of those to the property. It is not a violation of completeness and it is not a demonstration of it either: the comparison was not made, and a result that rests on a comparison nobody made is conditional (MUST-T11-12).

Checkpoint claims An epoch checkpoint MUST be COSE_Sign1-signed with the header profile in and MUST bind all of the following (MUST-T11-1): epoch, startMs, endMs, receiptCount, chainHeadHash, totals, and prevCheckpointHash. The checkpoint window is half-open [startMs, endMs) (MUST-T11-7). receiptCount MUST equal the number of receipts (settled and aborted) whose timestampMs falls in that window. chainHeadHash MUST equal receiptHash of the last receipt in that window, or null if the window is empty (MUST-T11-2). Where totals is present it MUST sum only receipts with outcome = settled; the one permitted absence is the signed redaction below. An issuer that publishes a checkpoint without its totals MUST encode totals as null in the signed payload (MUST-T11-12). An empty map is an honest zero for an empty window and is not a redaction. Because the redaction is inside the signature, it cannot be added to, or removed from, a checkpoint after signing. A verifier MUST NOT accept a redaction asserted anywhere but the signed payload (MUST-T11-13). A presentation-layer flag alongside a checkpoint is chosen by whoever presents it, which is the party under audit; honouring such a flag would let that party switch off the totals comparison for a checkpoint whose signed totals are wrong. The structural claims (epoch, startMs, endMs, receiptCount, chainHeadHash, prevCheckpointHash) MUST NOT be redacted: a checkpoint missing any of them does not decode, and a verifier MUST treat it as a failed checkpoint rather than as a redacted one.

Genesis and continuity The first checkpoint in a presented chain is the genesis checkpoint of that chain. Its prevCheckpointHash MUST be null. Epoch numbers MUST be consecutive integers. Adjacent windows MUST satisfy next.startMs = prev.endMs (MUST-T11-8). A later checkpoint that omits a prefix of earlier epochs (prefix deletion) is detectable only if an external witness (transparency log) has recorded the missing prefix (MUST-T11-9). Without that witness, T11 guarantees about suppression are conditional. -01 stated that dependency and stopped there: nothing in its verification algorithm read a transparency receipt, so the witness had no way to speak. -02 gave it one. A verifier that holds transparency receipts for the period under audit compares what the witness recorded against what the chain presented, and reports the difference under its own name (). A witness that holds a checkpoint the presented chain omits is not the same condition as a chain that leaves a gap in its own coverage, and the two MUST NOT be reported under one identifier (MUST-T11-11). The first says evidence is being withheld; the second says the evidence shown is incomplete. An operator who cannot tell them apart cannot tell an incomplete record from a concealed one.

The transparency witness A checkpoint registered with a Transparency Service is a Signed Statement whose payload is the checkpoint COSE object and whose content type is application/cedulon-checkpoint+cbor (). The service returns a receipt that binds the statement it recorded. A verifier MAY be given such receipts for the period under audit. It is a distinct input from the presented checkpoint chain, and supplying it is optional: a verifier given none performs the same steps, and reports the same findings, that it would if this input did not exist (MUST-T11-10). Supplying an empty set is not the same as supplying none. An empty set says a witness is configured and recorded nothing, which is itself reportable; absence says no witness was consulted. A receipt binds a statement hash, not a statement. The body is not carried by the receipt and a verifier will often not hold it. Two levels of checking follow from that, and they are not the same (MUST-T11-10). Every receipt MUST have its signature verified before it counts for anything. That is what establishes the hash as one the service signed for, and it is all that comparing recorded hashes against presented ones requires: the verifier computes the statement hash of each presented checkpoint itself. A receipt accompanied by the statement body carries more, and MUST be checked further before that body is relied on. The body's statement hash MUST equal the hash the receipt binds, and the body MUST itself verify as a checkpoint. A body that fails either check proves nothing and MUST be ignored rather than counted, while the receipt it came with remains usable for the hash comparison. Equivocation is the case that needs a body, because it compares claims rather than hashes. What such a receipt establishes, and what it does not, is worth stating plainly. It establishes that the service signed for that statement. Whether the statement is a member of an append-only log, and whether that log has ever equivocated, are properties of the service and its own proofs, not of Cedulon. A verifier that treats a signed receipt as proof of log membership is claiming more than the receipt carries.

Verification algorithm A verifier MUST perform all of these steps and MUST report every finding they produce (MUST-T10-1, MUST-T11-2). They are numbered for reference, not to require an evaluation order: no step short-circuits another, and an implementation may evaluate them in any order that produces the same set of findings. One data dependency is worth naming, because "any order" read naively would break it. Step 15 decides which transparency receipts, and which statement bodies, survive checking. Steps 14 and 16 consume what survives. An implementation that ran step 14 against unchecked bodies, or step 16 against unchecked receipts, would not produce the same set of findings, so that order is not among the permitted ones. Nothing else in this list feeds another step. When a step names an identifier in backticks, that identifier SHOULD be used for the condition in diagnostic output. The normative requirement is the behaviour: report the condition, identified by the ref or other handle given in the step. The identifiers are not an interoperability surface.

Establish the subject of the audit. When an extract is supplied, the settlement records it carries are the ones reconciled; a settlement list from any other source MUST NOT be substituted for them (MUST-T10-12). If the caller supplies both and they differ, the verifier MUST report that the caller-supplied list disagrees with the extract, and MUST still reconcile the extract. The identifier extract-settlement-mismatch SHOULD be used for this condition in diagnostic output.
Verify the extract signature against the out-of-band rail key (MUST-T10-8, MUST-T10-9). If no key is pinned, the verifier MUST treat the completeness guarantee as conditional (MUST-T10-7). The identifier unauthenticated-extract SHOULD be used for this condition in diagnostic output. If a key is pinned and cannot be decoded, the verifier MUST report that the pinned key is unreadable. The identifier trust-key-unreadable SHOULD be used for this condition. If a key is pinned and the signature does not verify against it, or verifies against a different key, the verifier MUST report that the extract is not signed by the pinned key. The identifier extract-key-mismatch SHOULD be used for this condition. A finding that puts the extract itself in doubt MUST prevent an unconditional guarantee.
Check scope. The verifier MUST report each settlement record whose timestampMs falls outside the declared window, identified by that record's ref (MUST-T10-10). When the verifier states an expected account, rail, or window, it MUST report an extract that does not cover it (MUST-T10-11). The identifier extract-scope-mismatch SHOULD be used for both conditions. If the verifier stated no period, it MUST treat the guarantee as conditional (MUST-T10-15). The identifier unstated-audit-window SHOULD be used for this condition.
Decode each Spend Receipt COSE_Sign1. Reject if Ed25519 verify fails, if kid does not match the configured issuer key, if the content type is not the receipt type, or if the decoded claim map does not match the presented claims (MUST-T4-2, MUST-T4-8). Then apply the issuer root (): verify the signature against the out-of-band issuer key rather than the key the receipt carries. A receipt that does not verify against it is reported and is excluded from the reconciliation that follows, so the settlement it names is still reported as uncovered in step 8 (MUST-T4-9, MUST-T4-10). With no issuer key pinned the verifier makes no such distinction and reports that it did not; where any receipt or checkpoint was presented, the guarantee is conditional on the terms in . Where a countersignature is present, governs what it establishes (MUST-T4-13, MUST-T4-14). Where a Trade Manifest is presented, governs it (MUST-T4-15): with no publisher key pinned the verifier reports unauthenticated-manifest and the guarantee is conditional; with a pin that cannot be read, trust-key-unreadable; with a pin the manifest does not answer to, manifest-key-mismatch; and with a manifest that no presented receipt references, manifest-covers-no-receipt (MUST-T4-17). A receipt that names the manifest but departs from its amount, currency or expiry is reported as manifest-terms-mismatch and fails the audit (MUST-T8-9). An audit presented with no Trade Manifest is not made conditional by this step.
Scope the receipts. When an extract is supplied, only receipts whose timestampMs falls in the extract's declared window are reconciled against it (MUST-T10-16). A receipt outside that window is not a completeness failure against this extract; auditing a longer period requires extracts that cover it. Receipts remain subject to every other check regardless of window.
Walk receipts in issuer order. The first prevReceiptHash MUST be null. Each later prevReceiptHash MUST equal receiptHash of the previous receipt. A miss MUST be reported as a break in the receipt chain. The identifier receipt-chain-break SHOULD be used for this condition.
Index settled receipts and extract records by ref. A ref that appears more than once on either side MUST be reported as a repeated reference (MUST-T10-6). The identifier duplicate-ref SHOULD be used for this condition.
For each ref that appears exactly once on each side, require a one-to-one match on ref AND amount AND currency (MUST-T10-1). Amount or currency mismatch MUST be reported as a settlement that does not match its receipt, identified by that ref. The identifier settlement-mismatch SHOULD be used for this condition. A settlement with no receipt MUST be reported as lacking a receipt, identified by its ref (MUST-T10-2). The identifier settlement-without-receipt SHOULD be used for this condition. A settled receipt with no extract row MUST be reported as a completeness failure (MUST-T10-3). The identifier receipt-without-settlement SHOULD be used for this condition. A settled receipt with a null rail ref MUST be reported as settled without a rail reference. The identifier settled-without-ref SHOULD be used for this condition.
A ref already reported as repeating MUST still be reconciled by amount rather than dropped from the comparison (MUST-T10-13). For each currency under that ref, compare the total settled against the total receipted. A settled total that exceeds the receipted total MUST be reported as a settlement lacking a receipt, and the finding MUST state the unaccounted amount. The identifier settlement-without-receipt SHOULD be used for this condition. A settled total that is less than the receipted total MUST be reported as a settlement that does not match its receipt, identified by that ref. The identifier settlement-mismatch SHOULD be used for this condition. An amount on that repeating ref that cannot be parsed as an integer MUST be reported without abandoning the audit; the identifier malformed-amount SHOULD be used for this condition. A verifier MUST still report findings for the remaining records.
Aborted receipts are not matched to extract rows and are not added to totals.
Decode each checkpoint. Reject a failed signature, and reject a kid that does not match the key obtained for the checkpoint issuer, on the same terms as a receipt (MUST-T4-8). Require receiptCount, chainHeadHash, and totals to match the receipts in [startMs, endMs) as defined above (MUST-T11-2). The identifier checkpoint-total-mismatch SHOULD be used for a failed signature, a wrong receiptCount, or totals that disagree, and checkpoint-head-mismatch for a chainHeadHash that is not the last in-window receipt. If the signed totals is null, the verifier cannot perform the totals comparison for that checkpoint. It MUST report that the comparison was skipped and MUST treat the completeness guarantee as conditional; the absence of a comparison is not a passed comparison (MUST-T11-12). The identifier checkpoint-totals-redacted SHOULD be used for this condition. receiptCount and chainHeadHash MUST still be checked. A checkpoint that fails verification MUST NOT be treated as redacted, whatever it claims about its own totals.
Every chained receipt MUST fall in exactly one checkpoint window. A gap or double count MUST be reported as a window coverage failure (MUST-T11-7, MUST-T11-8). The identifier window-coverage SHOULD be used for this condition.
Walk checkpoints in epoch order. prevCheckpointHash MUST equal the SHA-256 of the previous checkpoint COSE bytes, or null for genesis (MUST-T11-4). The identifier checkpoint-total-mismatch SHOULD be used for a broken chain, which is the fourth condition its table row names.
If two successfully verified checkpoints share an epoch number and have different hashes, the verifier MUST report equivocation (MUST-T11-3). The identifier equivocation SHOULD be used for this condition. The checkpoints compared here are those presented together with any carried by verified transparency receipts (step 15). Comparing only the presented chain cannot raise this finding: MUST-T11-8, applied in step 12, requires that chain's epochs to be consecutive, so no two of its members share an epoch. A copy recorded by a witness is where the second one is found.
If transparency receipts were supplied, verify them against the out-of-band transparency key (); receipts that cannot be checked that way are not evidence in either direction and the verifier reports that it left them out (MUST-T11-15). Discard any whose signature fails (MUST-T11-10). A surviving receipt whose statement body verifies against the issuer root is a statement that issuer published; one that does not is another party's, and is not this issuer equivocating (MUST-T11-16). The survivors are the recorded statement hashes used in step 16. Where a receipt also carries the statement body, discard that body unless its statement hash equals the one the receipt binds and it verifies as a checkpoint; the surviving bodies are what step 14 compares. Discarding a body does not discard its receipt.
Compare the surviving witness records against the presented chain (MUST-T11-11). For each presented checkpoint with no surviving record, report that it is not anchored; the identifier checkpoint-not-anchored SHOULD be used. This is a warning: a witness may have been configured after the checkpoint was issued, and an operator's own gap is not evidence of concealment. For each surviving record whose statement is absent from the presented chain, report that a recorded checkpoint was withheld; the identifier checkpoint-withheld SHOULD be used. This is a finding and the audit MUST fail. A record that carries no statement body cannot say whose statement it binds, so it MUST NOT produce this finding; it is reported as an entry that could not be attributed and makes the guarantee conditional, because a real withholding must not be buried by removing the body (MUST-T11-17). Such a record still establishes anchoring in the first half of this step: proving that a checkpoint was logged needs only its hash. The verifier MUST NOT report a withheld checkpoint as a window coverage failure (MUST-T11-11).
If any finding remains that is not a warning (a warning is a condition that only makes the completeness guarantee conditional), the audit MUST fail (MUST-T10-4).

Finding codes The identifiers below are for diagnostic output. They are not an interoperability surface. A finding object that can be carried on the wire is outside the scope of this document and may be defined later. Two implementations interoperate when they accept the same inputs and fail or warn on the same conditions, not when they print the same strings. A condition that makes the audit fail is a finding. A condition that only makes the completeness guarantee conditional is a warning. Warnings MUST still appear in operator-facing output (MUST-T10-14).

Code	Effect	Meaning
settlement-without-receipt	audit fails	Extract row has no matching settled receipt, or a repeating `ref` settled more than it receipted
receipt-without-settlement	audit fails	Settled receipt ref is not on the extract
settlement-mismatch	audit fails	Same `ref`, different amount or currency, including a repeating `ref` that settled less than it receipted
duplicate-ref	audit fails	Ref appears more than once on one side
settled-without-ref	audit fails	`outcome` is settled and `x402PaymentRef` is null
receipt-chain-break	audit fails	Signature or `prevReceiptHash` failed
checkpoint-total-mismatch	audit fails	Totals, count, signature, or checkpoint chain failed
checkpoint-head-mismatch	audit fails	`chainHeadHash` is not the last in-window receipt
equivocation	audit fails	Two distinct hashes for one epoch
window-coverage	audit fails	Gap, overlap, or non-adjacent / non-consecutive windows
unauthenticated-extract	guarantee conditional	No pinned rail key, or the extract has no verifiable signature
extract-key-mismatch	audit fails	Extract is signed by a key other than the pinned rail key, or does not verify against it
trust-key-unreadable	audit fails	The pinned rail key could not be decoded; the verifier's configuration is at fault
issuer-key-mismatch	audit fails	An object is signed by a key other than the pinned issuer key, so it is not coverage for anything it names
countersign-key-mismatch	audit fails	A countersignature is by a key other than the one pinned for that payee
countersign-missing	conditional	A payee key is pinned and a settled receipt for that payee carries no countersignature
unauthenticated-issuer	conditional	No verifier-supplied issuer key and at least one receipt or checkpoint presented; those objects were checked against the keys they carry
unauthenticated-witness	conditional	No verifier-supplied transparency key; inclusion receipts were left out of the comparison
unauthenticated-countersigner	conditional	No verifier-supplied payee key; a countersignature is present but proves no approval
unauthenticated-manifest	conditional	No verifier-supplied manifest key and a Trade Manifest was presented; it was checked against the key it carries. An audit presented with no Trade Manifest is not this condition
manifest-key-mismatch	audit fails	A presented Trade Manifest is signed by a key other than the pinned publisher key, or does not verify against it
manifest-covers-no-receipt	conditional	A presented Trade Manifest is referenced by no presented receipt, including aborted ones and those outside the extract window; the terms were attributed but no receipt names them
manifest-terms-mismatch	audit fails	A presented receipt names this Trade Manifest but its amount, currency or settlement time departs from the manifest; a gate applying `MUST-T8-2` and `MUST-T3-3` would have refused the payment
witness-entry-unattributable	conditional	The witness holds a statement this chain does not present, carrying no body to say whose it is
extract-scope-mismatch	audit fails	A record falls outside the declared window, or the extract does not cover the expected account, rail, or window
extract-settlement-mismatch	audit fails	A caller-supplied settlement list disagrees with the extract; the extract is authoritative
malformed-amount	audit fails	An amount on a `ref` already reported as repeating that could not be parsed as an integer
unstated-audit-window	guarantee conditional	The verifier stated no period, so the extract defined its own
countersign-bad	audit fails	Present payee countersignature failed verify
checkpoint-withheld	audit fails	A verified transparency receipt binds a checkpoint the presented chain does not contain
checkpoint-not-anchored	guarantee conditional	A witness was supplied and holds no verified receipt for this checkpoint
checkpoint-totals-redacted	guarantee conditional	The checkpoint was signed with `totals` null, so the totals comparison could not be made

A finding that puts the extract itself in doubt (extract-key-mismatch, trust-key-unreadable, extract-scope-mismatch, or extract-settlement-mismatch) MUST also prevent an unconditional guarantee, not merely fail the audit. A finding that puts a presented Trade Manifest in doubt (manifest-key-mismatch, or trust-key-unreadable on the manifest pin) does the same. An unconditional guarantee therefore requires all of: an extract, a pinned rail key the extract's signature verifies against, a stated period the extract covers, an issuer root for whatever receipts and checkpoints are presented, a manifest root for whatever Trade Manifest is presented, no finding that puts the extract in doubt, and no warning that withholds part of the comparison. A checkpoint whose totals were signed as withheld (checkpoint-totals-redacted) removes a comparison the guarantee rests on, and a presented checkpoint a supplied witness does not hold (checkpoint-not-anchored) leaves part of the chain unwitnessed. Either one makes the result conditional. Anything less than the whole list is conditional, and the report MUST say so. An implementation MUST make the guarantee and any warnings visible in whatever human-readable audit report it produces under this document, not only in a returned structure (MUST-T10-14). A report that says the books balance while withholding that the balance is conditional invites the reader to take a conditional result for an unconditional one. Checkpoints SHOULD be registered with a Transparency Service (SHOULD-T11-5). A test deployment MAY use an in-process append-only log as the witness (MAY-T11-6). Cedulon still MUST NOT take custody. The guarantee named above is about completeness against the extract, which is the subject of T10. It is not a claim that no checkpoint was suppressed. Suppression is the subject of T11, and a report MUST NOT be read as settling it when no witness was consulted: with no transparency receipts, the presented chain is self-consistent by construction and says nothing about what it left out (MUST-T11-9). A verifier that consulted a witness and found every presented checkpoint recorded, with nothing recorded that was not presented, has discharged T11 for the period those receipts cover, and for no longer.

Lifecycle

Manifest. Parties sign a Trade Manifest (optional for metered API spend; required for goods with acceptance criteria).
Policy check. The adapter submits a structured request to the PDP. Default is deny. An allow is a Decision Token (MUST-T6-4).
Payment. On allow, the adapter performs the x402 (or other rail) exchange using exactly the decision fields (MUST-T6-1). The Decision Token is consumed (MUST-T6-2). A reused nonce is denied (MUST-T3-1, MUST-T3-2). A tampered or expired token is denied (MUST-T6-5).
Receipt. The Receipt Issuer signs a Spend Receipt. Rail credentials MUST NOT appear in the receipt, logs, or tool results (MUST-T5-2, MUST-T7-1).
Dispute Evidence Bundle. If delivery bytes do not match the acceptance-criteria hash, an implementation MUST be able to emit a bundle of manifest + receipt + delivery hash (MUST-T8-3). The bundle MUST NOT be described as an arbitral award or escrow release (MUST-T8-4).

Policy Semantics Policy is default deny. The engine understands three families of rule:

Limit: maximum amount per payment; maximum cumulative amount per window (MUST-T2-2).
Velocity: maximum number of allowed payments per window (MUST-T2-1).
Scope: optional allow-lists for payee, currency, and tool name.

Fail-closed: missing engine, crash, or exception yields deny (MUST-T2-3). Implementations SHOULD emit stable reason codes (SHOULD-T2-5). Decision tokens SHOULD expire after a short TTL (SHOULD-T6-3). The agent-facing spend interface MUST invoke the PDP and MUST NOT expose a parallel ungated rail call to the model (MUST-T5-1).

SCITT Anchoring A Receipt Issuer or relying party MAY construct a SCITT Signed Statement whose payload is either the Spend Receipt COSE object or a privacy profile () and register it with a Transparency Service . The service returns a COSE receipt . Embedding that receipt yields a Transparent Statement. Cedulon does not define a new transparency algorithm. An epoch checkpoint MUST be registrable on the same terms (MUST-T11-14). Its Signed Statement carries the checkpoint COSE_Sign1 object as the payload and application/cedulon-checkpoint+cbor as the content type, which is among the media types asks to have registered and which, until then, is a placeholder like the rest. Nothing else about registration differs from a receipt. This is a short section for a requirement -01 was missing, and the omission mattered more than its length suggests. -01 asked for checkpoints to be registered (SHOULD-T11-5) while profiling only the receipt here, so the object carrying the suppression guarantee had no stated form to be registered in. Two implementations could follow -01 to the letter and register incomparable things.

Privacy Considerations A public transparency encoding MUST support omitting or hashing payer and payee identifiers and MUST support amount redaction or bucket encoding (MUST-T9-1). Implementations MUST NOT write government-ID numbers, payment-instrument PAN, or street address into a public statement (MUST-T9-2). Default public anchors SHOULD publish policyHash, manifestHash, receiptHash, and timestampMs rather than full claims (SHOULD-T9-3). A private auditor MAY receive an unredacted receipt out of band (MAY-T9-4). The paragraph above counts receipt fields. A checkpoint publishes something a receipt does not: a per-currency total for a whole window, which discloses trading volume even when every individual receipt is redacted (MUST-T9-5). -01 gave no rule for it, so an implementation could publish that total, or withhold it in a way no verifier could recognise, and neither reading contradicted the text. The rule is the one stated in : totals MAY be withheld by signing it as null (MUST-T11-12), and only that form counts as a redaction (MUST-T11-13). The structural claims are not redactable, because a verifier that cannot read the window or the chain head cannot check anything at all, and a checkpoint that hid them would be indistinguishable from a broken one. Withholding is honest and it is also a cost: a verifier that cannot recompute the totals says so, and the completeness guarantee for that window is conditional. A deployment that wants an unconditional result publishes the totals; a deployment that wants the volume private accepts a conditional one. What a deployment MUST NOT do is obtain the unconditional result while withholding the evidence for it.

Security Considerations This section is authoritative for the protocol requirements in this document. The companion repository file THREAT_MODEL.md is informative and MUST NOT be read as overriding this section. Requirement identifiers take the form KEYWORD-Tn-k, where KEYWORD is MUST, SHOULD, or MAY, n is the threat number in this section, and k is a sequence number within that threat. MUST-T8-custody is the custody prohibition under T8. The tables below define the requirement text those citations refer to.

T1: Prompt injection leads to unauthorized spend

ID	Requirement
MUST-T1-1	The PDP MUST decide from structured request fields and stored policy, not from model-generated prose.
MUST-T1-2	A spend that is not bound to a verified Trade Manifest MUST be marked `noManifest` on the Spend Receipt and MUST still be subject to limit, velocity, and scope policy.
SHOULD-T1-3	Hosts SHOULD require a human confirmation channel for first-use payees.
MAY-T1-4	An implementation MAY refuse all `noManifest` spend.

T2: Runaway agent (loop spend)

ID	Requirement
MUST-T2-1	Policy MUST express a maximum payment count per configured time window (velocity).
MUST-T2-2	Policy MUST express a maximum amount per payment and a maximum cumulative amount per window.
MUST-T2-3	If the PDP is unreachable, uninitialized, or throws during evaluation, the spend MUST be denied (fail-closed, default deny).
MUST-T2-4	A denied attempt MUST NOT increment the allowed-spend counters as if it had succeeded.
SHOULD-T2-5	Implementations SHOULD emit a stable reason code for velocity and limit denials.

T3: Replay of payment authority

ID	Requirement
MUST-T3-1	Every spend attempt that the PDP allows MUST include a nonce that the implementation has not accepted before.
MUST-T3-2	A second attempt that reuses a nonce MUST be denied.
MUST-T3-3	A Trade Manifest MUST carry an expiry; a spend against an expired manifest MUST be denied.
MUST-T3-4	A PDP allow decision MUST be bound to a hash of the request fields it evaluated and MUST be single-use.
SHOULD-T3-5	Nonce stores SHOULD persist across process restart when the deployment is not a test fixture.

T4: Receipt forgery or repudiation

ID	Requirement
MUST-T4-1	A Spend Receipt MUST be signed by the Receipt Issuer over a canonical encoding of its claims.
MUST-T4-2	Verifiers MUST reject a receipt whose signature does not validate or whose canonical bytes do not match the signed payload.
MUST-T4-3	A Spend Receipt MUST include `payer`, `payee`, `amount`, `currency`, `policyHash`, `timestampMs`, and `nonce`.
MUST-T4-4	A Spend Receipt MUST include `manifestHash` or an explicit `noManifest` flag, never an ambiguous empty hash. Empty optional values are CBOR null; labels are never absent.
SHOULD-T4-5	Receipts SHOULD form a hash chain (`prevReceiptHash`) so omission is detectable within one issuer stream.
MAY-T4-6	Parties MAY register the signed receipt as a SCITT statement to obtain a COSE receipt.
MUST-T4-7	A Spend Receipt MUST include `outcome` (`settled` or `aborted`). A settled receipt MUST have a non-null rail ref. Aborted receipts MUST NOT enter checkpoint totals.
MUST-T4-8	COSE_Sign1 protected headers MUST use alg -19 (Ed25519), a mandatory `kid`, and a payload-specific content type. Verifiers MUST reject a `kid` that does not match the configured issuer key.
MUST-T4-9	A verifier MUST obtain the issuer public key out of band and MUST verify Spend Receipt and epoch checkpoint signatures against that key, not against a key the object carries. A verifier without such a key that is presented with any Spend Receipt or epoch checkpoint MUST report the completeness guarantee as conditional. An audit presented with neither rests on the extract alone and is not made conditional by this requirement.
MUST-T4-10	A receipt that does not verify against the pinned issuer key MUST NOT count as coverage for the settlement it names, and that settlement MUST still be reported as uncovered. Reporting the mismatch is not sufficient on its own.
MUST-T4-11	Pinned issuer keys MUST be compared by SubjectPublicKeyInfo DER encoding. A pinned key that cannot be decoded MUST be reported as a verifier configuration fault rather than as a mismatch, and where no pinned key decodes, the verifier MUST NOT fall back to the keys the objects carry.
MUST-T4-12	A verifier MUST accept an issuer, publisher, witness, or rail root comprising more than one key, so that a key rotation inside the audited window does not require it to abandon pinning.
MUST-T4-13	A payee countersignature MUST NOT be treated as evidence of payee approval unless it verifies against a payee key the verifier obtained out of band.
MUST-T4-14	Where a verifier has pinned a key for a payee, a settled receipt naming that payee and carrying no countersignature MUST be reported, so that deleting the evidence does not delete the question.
MUST-T4-15	A verifier that is presented with a Trade Manifest MUST obtain the publisher public key out of band and MUST verify the manifest signature against that key, not against a key the manifest carries. A pin that cannot be read MUST be reported as `trust-key-unreadable`; a readable pin the manifest does not answer to MUST be reported as `manifest-key-mismatch` and MUST fail the audit. A verifier without such a key that is presented with a Trade Manifest MUST report the completeness guarantee as conditional. An audit presented with no Trade Manifest is not made conditional by this requirement.
MUST-T4-16	A policy decision point presented with a Trade Manifest it cannot verify against a key supplied out of band MUST refuse the payment. Settling and reporting the doubt afterwards is not available to it: the receipt carries the manifest hash as terms the named party agreed to.
MUST-T4-17	A verifier presented with a Trade Manifest MUST compare the manifest hash against the `manifestHash` of the receipts presented to the audit, including aborted ones and before any extract window is applied, and MUST report a presented manifest that no presented receipt references. Verifying who published the terms does not establish that any receipt names them. An audit presented with no Trade Manifest is not made conditional by this requirement.

T5: Policy bypass via direct rail access

ID	Requirement
MUST-T5-1	The agent-facing spend interface MUST invoke the PDP and MUST NOT expose a parallel ungated rail call to the model.
MUST-T5-2	Rail credentials, wallet handles, and facilitator tokens MUST NOT be placed in tool results or prompts.
SHOULD-T5-3	Hosts SHOULD run the PDP and signing keys in a process the model runtime cannot write.
MAY-T5-4	A deployment MAY use OS or hardware isolation between the model and the PDP.

T6: TOCTOU between policy check and payment

ID	Requirement
MUST-T6-1	Payment settlement MUST use the same six `requestHash` fields the PDP evaluated: amount, currency, payee, tool, nonce, and `manifestHash`.
MUST-T6-2	An allow decision MUST be consumed on the first settlement attempt, success or fail-closed abort, and MUST NOT authorize a later different request.
SHOULD-T6-3	Implementations SHOULD treat a decision older than a short TTL as expired.
MUST-T6-4	An allow Decision Token MUST be COSE_Sign1 with CWT private-use labels -70301..-70305 (`requestHash`, `policyHash`, `expiryMs`, `nonce`, `singleUseId`) and content type `application/cedulon-decision+cbor`.
MUST-T6-5	A party that accepts a Decision Token MUST reject a failed signature, a `kid` or content-type mismatch, a claim-map mismatch, or an expired `expiryMs`.
MUST-T6-6	A consumer of a Decision Token MUST verify it against its own issuing key and MUST NOT accept a token it cannot check that way. The consumer issued the token, so asking the token which key to check it against is a question that answers itself.

T7: Signing-key leakage

ID	Requirement
MUST-T7-1	Secret key material MUST NOT appear in receipts, checkpoints, manifests, decision tokens, logs, or example output.
MUST-T7-2	Example and test keys MUST be generated at runtime or stored as clearly fake fixtures, never as production secrets.
SHOULD-T7-3	Production deployments SHOULD use an HSM or OS key store and SHOULD rotate keys.
MAY-T7-4	Implementations MAY encrypt keys at rest.
MUST-T7-5	An implementation that stores a signing key in the clear MUST report the protection it actually obtained, measured from the stored object rather than derived from the platform. A mount that ignores filesystem permissions accepts the call and protects nothing.
MUST-T7-6	A writable directory anywhere on the path to a stored signing key makes the file permission moot, and a symbolic link on that path hands the destination to whoever placed it. An implementation MUST refuse both rather than report the key as protected.

T8: Counterparty price gouging or defective delivery

ID	Requirement
MUST-T8-1	A Trade Manifest MUST bind goods or service description, price, currency, acceptance-criteria hash, cancel condition, and expiry.
MUST-T8-2	A spend bound to a manifest MUST be denied if the requested amount or currency differs from the manifest.
MUST-T8-3	If delivery bytes do not hash to the acceptance-criteria hash, the implementation MUST be able to produce a Dispute Evidence Bundle containing the manifest, the spend receipt, and the delivery hash.
MUST-T8-4	The Dispute Evidence Bundle MUST NOT be described as an arbitral award or escrow release.
MUST-T8-7	`manifestHash` MUST be the SHA-256 of the signed Trade Manifest COSE bytes and MUST NOT include the issuer public key encoding.
SHOULD-T8-5	Manifests SHOULD reference an AP2 mandate hash when one exists.
MAY-T8-6	Parties MAY add an optional escrow actor as a third-party role interface; this project MUST NOT implement custody.
MUST-T8-custody	Implementations of this specification MUST NOT take custody of funds or operate escrow.
MUST-T8-8	If a payee countersignature is present, a verifier MUST reject it when the signature fails, when `kid` or content type does not match the configured payee key, or when the payload is not the issuer COSE_Sign1 bytes.
MUST-T8-9	A verifier presented with a Trade Manifest MUST compare the amount, currency and settlement time of every presented receipt that names it against the manifest amount, currency and expiry, and MUST report a receipt that departs from them. `MUST-T8-2` and `MUST-T3-3` bind the gate; an audit reads the record after the gate is gone, so without this the receipt can carry the hash of terms it breaks. Receipts that do not name the manifest are not measured against it.
MAY-T8-9	A payee MAY attach a detached COSE_Sign1 countersignature over the issuer receipt bytes. Absence MUST NOT invalidate the issuer receipt.

MUST-T8-custody Implementations of this specification MUST NOT take custody of funds or operate escrow. See also .

T9: PII leakage into the transparency log See also .

ID	Requirement
MUST-T9-1	A transparency encoding MUST support omitting or hashing payer/payee identifiers and MUST support amount redaction or range/bucket encoding.
MUST-T9-2	Implementations MUST NOT write raw government-ID, payment-instrument PAN, or street address fields into a public transparency statement.
SHOULD-T9-3	Default public anchors SHOULD publish `policyHash`, `manifestHash`, `receiptHash`, and timestamp rather than full claim sets.
MAY-T9-4	A private auditor MAY receive an unredacted receipt out of band.
MUST-T9-5	A checkpoint discloses a per-currency window total, which the receipt-field rules above do not cover. Withholding it is governed by MUST-T11-12 and MUST-T11-13: null in the signed payload, and no other form of redaction honoured.

T10: Secret spend via rail bypass See .

ID	Requirement
MUST-T10-1	A verifier MUST match each extract settlement to a settled receipt on `ref` AND `amount` AND `currency`.
MUST-T10-2	A settlement with no matching receipt MUST be reported as a completeness failure identified by that settlement `ref`.
MUST-T10-3	A settled Spend Receipt whose `x402PaymentRef` is not on the extract MUST be reported as a completeness failure.
MUST-T10-4	An audit that has any fail-severity completeness finding MUST fail (non-zero status in the companion tool).
SHOULD-T10-5	Hosts SHOULD still apply T5 (no ungated rail in the model process). Completeness does not replace prevention.
MUST-T10-6	A `ref` that appears more than once among settled receipts or among extract rows MUST be reported as `duplicate-ref`.
MUST-T10-7	A verifier MUST obtain the extract from the rail or from a rail signature. With no pinned rail key, an unverifiable extract MUST be reported as `unauthenticated-extract` and makes the completeness guarantee conditional. With a pinned key, see MUST-T10-8: the extract MUST fail closed rather than warn.
MUST-T10-8	A verifier MUST obtain the rail public key out of band and MUST verify the extract signature against that key, not against a key the extract carries. Without such a key the guarantee is conditional.
MUST-T10-9	Keys MUST be compared by SubjectPublicKeyInfo DER encoding rather than by any text encoding. A pinned key that cannot be decoded MUST be reported as `trust-key-unreadable`, not as a key mismatch.
MUST-T10-10	Every settlement record whose `timestampMs` falls outside the extract's declared window MUST be reported as `extract-scope-mismatch`, identified by that record's `ref`. This check MUST run whether or not a key is pinned.
MUST-T10-11	When the verifier states an expected account, rail, or window, an extract that does not cover it MUST fail closed as `extract-scope-mismatch`.
MUST-T10-12	When an extract is supplied, the records it carries are the subject of reconciliation. A settlement list from another source MUST NOT be substituted; a disagreeing list MUST be reported as `extract-settlement-mismatch`.
MUST-T10-13	A `ref` reported as `duplicate-ref` MUST still be reconciled by aggregate amount per currency, and a shortfall MUST state the unaccounted amount. An unparseable amount MUST be reported as `malformed-amount` without aborting the audit.
MUST-T10-14	An implementation MUST surface the guarantee and any warnings in any human-readable audit report it produces, not only in a returned structure.
MUST-T10-15	A verifier that has not stated the period under audit MUST emit `unstated-audit-window` and MUST treat the guarantee as conditional, because an unstated period leaves the extract free to define its own.
MUST-T10-16	When an extract is supplied, only receipts whose `timestampMs` falls in its declared window are reconciled against it. A receipt outside that window MUST NOT be reported as a completeness failure against that extract.

In MUST-T10-4, a completeness finding that makes the audit fail is distinct from a warning that only makes the guarantee conditional. The verification algorithm states that distinction by behaviour ().

T11: Checkpoint suppression or rollback

ID	Requirement
MUST-T11-1	An epoch checkpoint MUST be COSE-signed and MUST bind epoch number, time window, receipt count, chain-head hash, per-currency totals, and the previous checkpoint hash.
MUST-T11-2	Verifiers MUST reject a checkpoint whose signature fails, whose totals do not match settled receipts in the declared window, whose `receiptCount` is wrong, or whose `chainHeadHash` is not the last in-window receipt hash. Where the signed totals are null, MUST-T11-12 governs instead: there is no total to disagree with, the comparison is reported as skipped, and the count and chain-head checks still apply.
MUST-T11-3	Two verified checkpoints for the same epoch with different hashes MUST be reported as equivocation. The checkpoints compared are those presented together with those carried by verified transparency receipts; the presented chain alone cannot satisfy this requirement, because MUST-T11-8 makes its epochs consecutive.
MUST-T11-4	A broken checkpoint hash chain MUST fail verification.
SHOULD-T11-5	Checkpoints SHOULD be registered with a Transparency Service when one is configured.
MAY-T11-6	A test deployment MAY use an in-process append-only log as the witness.
MUST-T11-7	Checkpoint windows MUST be half-open `[startMs, endMs)`. Every chained receipt MUST fall in exactly one window.
MUST-T11-8	Presented checkpoint epochs MUST be consecutive and adjacent windows MUST meet at `endMs = next.startMs`.
MUST-T11-9	Prefix-deletion and suppression claims that go beyond the presented chain are conditional on an external transparency witness. A report MUST NOT present a completeness guarantee as settling suppression when no witness was consulted.
MUST-T11-10	Transparency receipts are an optional, separate input. A verifier given none MUST behave as it did without this input. A receipt MUST have its signature verified before it counts for anything. Where a receipt also carries the statement body, that body MUST NOT be relied on unless its statement hash equals the one the receipt binds and it verifies as a checkpoint; a discarded body does not discard its receipt.
MUST-T11-11	A verified receipt binding a checkpoint absent from the presented chain MUST be reported as a withheld checkpoint, and MUST NOT be reported as a window coverage failure. A presented checkpoint with no verified receipt, where a witness was supplied, MUST be reported and makes the guarantee conditional.
MUST-T11-12	Withheld checkpoint totals MUST be encoded as null in the signed payload. A verifier MUST report that the totals comparison was skipped and MUST treat the guarantee as conditional; `receiptCount` and `chainHeadHash` MUST still be checked.
MUST-T11-13	A redaction asserted outside the signed payload MUST NOT be honoured, and structural claims MUST NOT be redacted. A checkpoint that fails verification MUST NOT be treated as redacted.
MUST-T11-14	An epoch checkpoint MUST be registrable as a Signed Statement carrying the checkpoint COSE object with content type `application/cedulon-checkpoint+cbor`.
MUST-T11-15	A verifier MUST obtain the transparency service public key out of band and MUST verify inclusion receipts against it. Receipts that cannot be checked that way MUST NOT be used as evidence in either direction, and the verifier SHOULD report that they were left out.
MUST-T11-16	A statement held by a pinned log MUST additionally verify against the issuer root before it counts as something that issuer published, so that another user's epoch in a shared log is not read as equivocation by this issuer.
MUST-T11-17	An inclusion receipt carrying no statement body MUST NOT be used to report a withheld checkpoint, because it cannot say whose statement it binds. It MUST NOT be discarded silently either: a real withholding must not be buried by removing the body.

Issuer self-attestation:: A Receipt Issuer that also produces the only copy of the extract can omit settlements. Completeness holds only against an extract the verifier obtained from the rail or from a rail signature.
Key rotation and revocation:: kid identifies the verification key. This -03 does not specify a revocation list. Verifiers MUST pin the issuer keys they accept and MUST stop accepting a kid after an authenticated revocation signal.
Timestamp trust:: timestampMs is issuer-asserted. Window assignment uses that field. A lying issuer can slide a receipt between windows. External timestamping of receipts is out of scope for this revision; the checkpoint witness covers checkpoints, not receipt timestamps.
Collusion:: If the rail operator and the issuer collude, they can publish a matching extract and receipt set that hides a real-world settlement. Cedulon does not detect extract-external agreement.
Reversal, refund, and partial settlement:: State machines for reversal, refund, and partial settlement are out of scope for this revision.

Optional escrow role Parties MAY name an escrow actor in a Trade Manifest as a third-party role that holds funds under rules outside this protocol (MAY-T8-6). Implementations of this specification MUST NOT take custody or operate escrow (MUST-T8-custody).

T12: Settlement without a recorded receipt The threats above are about a counterparty, a rail or an attacker. This one is about the issuer's own implementation, and it produces exactly the condition the rest of this document exists to make detectable. An issuer that settles a payment, appends the receipt in memory and then persists its state has three steps where it could have two outcomes. If the write fails, the rail holds a settlement and the receipt exists nowhere durable. The next start reads a state that does not contain it, and the audit in reports settlement-without-receipt against an honest issuer that did everything its own policy asked. The evidence is missing because the issuer lost it, not because anyone hid it, and nothing in the report can tell those apart.

ID	Requirement
MUST-T12-1	An issuer MUST NOT complete a settlement whose receipt it cannot record durably. The ability to record MUST be established before value moves, not after.
MUST-T12-2	Where a settlement has been made and its record cannot be completed, the issuer MUST undo the settlement in every place it still controls: in memory, on the rail ledger it controls, and in any later write it has not yet issued. A refused payment MUST NOT consume the nonce or the payment allowance it never used.
MUST-T12-3	Two issuers MUST NOT share one durable state. An implementation that permits it MUST fail loudly rather than let one writer overwrite the other's receipt, and the failure MUST name what an operator can act on.
MUST-T12-4	Where a settlement has entered a rail the issuer does not control, and the local record cannot be completed, the outcome is indeterminate. The issuer MUST NOT treat the payment as reversed, and MUST NOT return the authority to spend, unless it holds authenticated evidence that the rail did not complete the settlement or that a reversing entry completed.

Ordering:: Settle-then-record is the natural order to write and the wrong one to ship. The record is what makes the settlement accountable, so the record is what has to be secured first (MUST-T12-1), and a settlement that cannot be recorded has to be undone everywhere the issuer still controls, including the nonce and the allowance it never used (MUST-T12-2). That undo is what the in-process RailLedger and the session tests measure. Once value has entered a rail the issuer does not control, a local snapshot cannot retract it. Persistence failing after that point leaves the outcome indeterminate; authority is not returned without authenticated evidence that the rail did not complete the settlement or that a reversing entry did (MUST-T12-4).
Recovery:: A durable-state conflict is not necessarily fatal, but it MUST NOT be silent, and an implementation that refuses every subsequent write without offering a way back has turned a recoverable condition into an outage. The reason reported has to separate the cases an operator would act on differently (MUST-T12-3): a write that conflicted with another writer, a write that failed, and a state another process is holding. A single opaque failure leaves the operator to guess which of those happened.
Observability:: A violation of this threat is not visible in the evidence a verifier receives. The audit sees a settlement with no receipt and reports settlement-without-receipt, which is the same finding an adversary would produce, and nothing in the extract or the receipt set distinguishes an issuer that lost the evidence from one that hid it. That is why the requirements here fall on the issuer rather than on the verifier, and why an operator-facing reason is required rather than optional.

IANA Considerations CWT claim labels need no assignment: this profile uses private-use integer labels below -65536. Media types do. This document defines five and makes their values a normative check inside a protected COSE header (MUST-T4-8, MUST-T6-5), so they cannot stay unregistered while that check stands. If this work is taken up, the following should be registered in the Standards Tree per :

Media type	Carries
application/cedulon-receipt+cbor	Spend Receipt claim set
application/cedulon-checkpoint+cbor	Epoch checkpoint claim set
application/cedulon-manifest+cbor	Trade Manifest claim set
application/cedulon-decision+cbor	Decision Token claim set
application/cedulon-countersign+cbor	Payee countersignature

For each: encoding is binary CBOR ; security considerations are those in ; the change controller would be the IETF; the contact is the author of this document. Until registration, an implementation outside a closed deployment should expect these names to change, and readers should treat them as placeholders rather than as stable identifiers.

Implementation Status This section is to be removed before publishing as an RFC. RFC 7942 note.

Implementation:: A companion implementation with a runnable verification suite at https://github.com/dogrucanemek-alt/cedulon. The code is a profile of this document, not a second specification. This -03 is not an IETF working-group item.
Maturity:: Research code by a single implementer. Three readers have run the code on their own machines against a pinned commit and reported figures matching the author's: two from a clean clone of the whole suite, one re-running the published reproduction. That is byte-stability across environments, not an independent implementation, and the distinction matters: the same code agreeing with itself on three machines rules out a local accident and nothing more. One reader reports an independent implementation of the Signed Statement identity, kept deliberately separate from this codebase; no independent implementation of the reconciliation algorithm is known to the author.
: The requirements added in this revision came out of five adversarial rounds against the implementation, each one asking a reviewer to break the code rather than to read it, with the reviewer barred from changing it. Four of those rounds found a defect inside the previous round's repair rather than in the original code, which is the reason this section does not describe the result as settled.
Coverage:: The receipt, checkpoint, extract, reconciliation, and verification algorithm are implemented, including the transparency witness input, the withheld and not-anchored conditions, and signed totals redaction. Requirements added in this revision are implemented and covered by a red-then-green case before appearing in this text, except MUST-T12-4, which is specified and not executed, and MUST-T4-17 and MUST-T8-9, which are implemented in this tree and not in the published 0.4.0 packages (see the note on distribution below). Some of those cases need POSIX file modes or symbolic links and assert only there; on Windows they return early. That leaves the symbolic-link protections and the undo after a failed write unexercised by a passing suite on Windows, while the protection report and the refusal to settle without a durable record are checked on both. An independent runner reported the distinction back from a Linux run after this text first claimed otherwise, and this is the second correction to the same sentence. The witness used in the suite is the in-process append-only log that MAY-T11-6 permits; the implementation has not been run against a deployed Transparency Service, and it treats a receipt as a signature over a statement, not as a proof of log membership. The escrow role, reversal, refund, and partial settlement are not implemented.
Licensing:: Apache-2.0.
Contact:: The author of this document.
Experience:: Readers of -00, -01 and -02 have reported defects in each, and every revision has been driven by what they found rather than by a plan. -01 fixed a bypass of the completeness claim and a gap about which key an extract is checked against. -02 repaired a defect reported against -01 and independently confirmed by a second reader: the object carrying the T11 guarantee was neither profiled for registration nor read during verification.
: The defect behind this revision was reported against the posted -02. A reader asked whether the profile should accept a pinned witness key and report an absent or mismatched pin explicitly. -02 Section 6.2 already required a verifier to obtain the public key from an authenticated channel and to reject a kid that does not match that key. What it did not carry was the verification algorithm, the separate root inputs, and the error semantics. Following the same question into the implementation found the omission for the Spend Receipt, the epoch checkpoint and the Decision Token. T12 came from neither reader nor adversary, and not from the rounds of trying to break the implementation either: it was found while writing the task for one of them, in the ordering the implementation itself used, which produced against the issuer the one condition this document exists to make detectable.

Note on distribution: the requirements this revision adds that are in a published package are in the published @cedulon packages at version 0.4.0, not only in the repository, with the exceptions named below. A reader can check a claim against an installed package rather than against a working tree. That order is deliberate: -00 described requirements that its published package did not yet carry, a reader found the discrepancy, and this document does not repeat it. Versions 0.2.x and earlier predate everything in this revision. 0.3.0 predated the manifest root and the T12 bound. It carried three defects that only appear away from the platform it was written on, which an independent runner found by taking up a standing invitation to break it. A directory that could not be written refused the lock before the record and left the refusal as an uncaught exception rather than the reason this document requires; the case for a symbolic link on the path used a call that does not exist in the module system the package declares, so it never reached its assertion; and repairing that revealed a fourth defect, that the state fingerprint was read before the path was checked, so a replaced path was reported as a conflicting writer rather than as a hijacked destination. 0.3.1 closes all four. The manifest root (MUST-T4-15) and the gate's refusal to settle against a manifest it cannot attribute (MUST-T4-16) were published as 0.4.0 rather than as a patch: the gate had been answering 200 to an unattributable manifest and writing that manifest's hash into the receipt, and refusing it is a change in behaviour that a version number ought to announce. MUST-T4-17, MUST-T8-9 and MUST-T12-4 are the exceptions in this revision. The same independent runner who took up the invitation against 0.4.0 reported that attributing a manifest was not the same as establishing that anything in the window was spent under it, which is the distinction that MUST-T4-17 now draws. A reader of this document then observed that the distinction survives one step further out: a receipt can name the manifest and still depart from its amount, currency or expiry, which is what MUST-T8-9 closes. Both are implemented and tested in this tree and are not a published npm release until the author says so; a reader checking either against an installed 0.4.0 will not find it. MUST-T12-4 is specified, not executed: the suite and the published server only drive the in-process RailLedger. There is no authenticated external-rail path in this tree, so the indeterminate outcome and the rule that forbids returning authority without evidence have no red-then-green case. A reader checking MUST-T12-4 against an installed 0.4.0 will not find it.

Changes from -02 This -03 has two subjects. The first is that -02 stated a rule for signed objects and left the implementation and the verification algorithm without a counterpart for every object the rule applied to. The second is T12, which no reader reported and which is not about an adversary at all. -02 Section 6.2 already required a verifier to obtain the public key from an authenticated channel (a preconfigured issuer set, a directory, or a transparency statement) and to reject a message whose kid does not match that key (MUST-T4-8). That rule was general. What -02 did not carry was the verification algorithm that applies it to each signed object, the separate root inputs a verifier supplies out of band, and the error semantics that name a missing, unreadable, or mismatched pin. A companion implementation that followed the algorithm it had, rather than that prose, still checked a Spend Receipt, an epoch checkpoint, a Decision Token and a transparency inclusion receipt against the key travelling inside them. A receipt like that silences the settlement-without-receipt finding for the settlement it names. The completeness property is then computed over evidence that answers to nobody. states the missing algorithm and the error semantics, and the verification algorithm now applies them. A reader raised the inclusion-receipt half of this against the posted -02 after checking the archived text against the implementation commit. Following it into the code turned up the other three, along with two conditions that were not about keys at all. The first is that reporting a mismatch is not enough. An implementation that names a foreign key and then still lets the receipt match its settlement has described the attack in its output while concluding that the books balance. MUST-T4-10 now requires the settlement to stay reported. The second is that an expectation which only fires when the evidence is present can be cancelled by deleting the evidence. A pinned payee key with no countersignature to check was silence, so removing a countersignature removed the question with it. MUST-T4-14 closes that, and MUST-T11-17 closes the same shape in the witness: an inclusion receipt with its body stripped off can no longer bury a withholding, while still not being allowed to accuse anyone. T12 is new and is not about an adversary. An issuer that settles, appends the receipt in memory and then writes its state will, when that write fails, leave the rail holding a settlement whose receipt exists nowhere durable. Restarted, it reports settlement-without-receipt against itself. The condition this document exists to make detectable was reachable through the implementation's own ordering, and no requirement in -02 said otherwise. The same first subject carries five requirements this section has not named so far, all of them stated in . A pinned key that cannot be decoded MUST NOT fall back to the keys the objects carry (MUST-T4-11). An issuer root may comprise more than one key, and a verifier must accept one that does, so a rotation inside the audited window does not force it to choose between findings against honest receipts and abandoning the pin (MUST-T4-12); the same set-of-keys acceptance applies to a publisher, witness, or rail pin. The same out-of-band rule reaches the payee countersignature (MUST-T4-13), the transparency witness (MUST-T11-15 and MUST-T11-16), the Decision Token, whose consumer issued it and therefore already holds the key to check it with (MUST-T6-6), and a presented Trade Manifest (MUST-T4-15). Two requirements belong to neither subject. MUST-T7-5 and MUST-T7-6 come from measuring the protection a stored signing key actually has instead of deriving it from the platform. A mount that ignores filesystem permissions accepts the call and protects nothing, and a writable directory or a symbolic link anywhere on the path makes the file permission moot. They are stated because the implementation reported protection it did not have. Two things are stated here that -02 got right and this revision keeps unchanged: the extract rule itself, and the treatment of a pinned key the verifier cannot decode. What changed is their reach, and the change is not backward compatible. Where a verifier pinned the rail key, supplied no issuer key, and was presented with receipts or checkpoints, -02 reported the guarantee as unconditional and this revision reports it as conditional. Nothing about the evidence changed; what changed is that the guarantee now says which questions were never asked. An audit presented with neither is unaffected, for the reason given in .

Changes from -01 -02 had one subject: the checkpoint, which carries the T11 guarantee against suppression and rollback, was not wired into anything that could discharge it. A reader of -01 set out the gap and a second reader confirmed it independently. Four things were wrong at once, and they were the same thing seen from four sides. The SCITT anchoring section profiled the Spend Receipt and not the checkpoint, so SHOULD-T11-5 asked for checkpoints to be registered without saying in what form (MUST-T11-14 now says). None of the steps of the verification algorithm read a transparency receipt, so a deployment could follow SHOULD-T11-5 to the letter and still have a verifier that never consulted the witness (steps 15 and 16 now do). MUST-T11-3, equivocation, could not fire at all: the only checkpoints compared were the presented ones, and MUST-T11-8 requires those to be consecutive, so no two of them can share an epoch. A second copy is found in a witness, and nothing brought one in. And the checkpoint binds a per-currency total for a window while the privacy section counted only receipt fields, so a window total could be published, or withheld, with no stated rule either way. The repairs are MUST-T11-10 through MUST-T11-14. The transparency receipt is a new optional input; supplying none leaves the verifier behaving exactly as in -01, which is deliberate, because the point of the witness is to add a claim that could not be made before, not to withdraw one that could. Two decisions inside those repairs are worth stating on their own, because a reader might reasonably have expected the other choice.

A recorded checkpoint the presented chain omits gets its own identifier, checkpoint-withheld, rather than being folded into window-coverage (MUST-T11-11). The reporter asked which way it should go. Coverage says the record shown is incomplete; a withheld checkpoint says the party under audit is holding a record it did not show. An operator who sees one identifier for both cannot tell an incomplete record from a concealed one, which is the distinction the whole threat is about.
A checkpoint with no receipt in a supplied witness is a warning rather than a failure. A witness may be configured after checkpoints have already been issued, and an operator's own gap in anchoring is not evidence that anything was concealed. The asymmetry is deliberate: what the witness holds and the chain does not is a finding, what the chain holds and the witness does not is a warning.

One change is not a repair of -01 but of the first attempt at this revision, and it is recorded because the failure is instructive. Signed totals were first made redactable through a field carried alongside the checkpoint rather than inside the COSE payload. Anything outside the signature is chosen by whoever presents the object, which here is the party under audit, so a checkpoint whose signed totals disagreed with its receipts could be re-presented as redacted and the mismatch went unreported. Redaction is now inside the signature and a redaction asserted anywhere else MUST be ignored (MUST-T11-12, MUST-T11-13). This was the same shape as the bypass -01 was written to close: a check that a party under audit could switch off. Reversal, refund, partial settlement, and the escrow role remain out of scope and are still expected later, with no date. -01 said the same and this revision does not improve on it. The reporters are named in the Acknowledgments.

Evolution and Future Work (Informative) This section is a direction, not a commitment. The structures below are reserved in name only. Normative wire formats, tests, and threat-model MUST lines for them belong in later revisions (-04 or later), written with the same discipline as this -03.

Re-attestation profile Algorithms retire. A Spend Receipt or checkpoint signed under Ed25519 today may need a later verifier that no longer accepts -19. A companion seed sketches re-attestation: register the original COSE bytes as a SCITT Signed Statement and have a current algorithm countersign or receipt them. The principle is that structures outlive ciphers. The first concrete example is the profile's own move from generic EdDSA (-8) to Ed25519 (-19) in .

Streaming reconciliation Epoch checkpoints in this document are batch windows. A later revision may define a continuous, second-scale profile in which the same completeness relation is evaluated as settlements arrive, without waiting for an epoch close. That work does not change the matching rules in this document, and it did not arrive in this revision either.

Generalization Payment is the special case that this -03 implements. The same completeness calculus (an authenticated extract of consumed units reconciled to signed receipts) can apply to other consumable resources such as compute, data, or energy. This document does not specify those profiles.

Informative Notes on Adjacent Protocols x402 uses HTTP 402 to negotiate stablecoin payment. AP2 uses signed mandates as verifiable credentials. Cedulon does not replace either protocol. Profiles built on HTTP Message Signatures authenticate bots; they are not a spend receipt. draft-bates-atp is a lineage neighbor. It does not define rail-extract completeness. draft-vauban-x402-stark-receipts specifies complementary x402 receipt-format variants that a Cedulon Spend Receipt MAY carry as a rail proof; it does not define rail-extract completeness. draft-schrock-ep-outcome-binding compares authorized action bytes to independently observed effects; it does not define rail-extract completeness. draft-marques-asqav-compliance-receipts profiles access-control action receipts (the broader Acta family includes ); it does not define rail-extract completeness. draft-hopley-x402-compliance-receipt records an admission-time compliance decision; it does not define rail-extract completeness.

References Normative References Key words for use in RFCs to Indicate Requirement Levels In many standards track documents several words are used to signify the requirements in the specification. These words are often capitalized. This document defines these words as they should be interpreted in IETF documents. This document specifies an Internet Best Current Practices for the Internet Community, and requests discussion and suggestions for improvements. US Secure Hash Algorithms (SHA and SHA-based HMAC and HKDF) Federal Information Processing Standard, FIPS Media Type Specifications and Registration Procedures This document defines procedures for the specification and registration of media types for use in HTTP, MIME, and other Internet protocols. This memo documents an Internet Best Current Practice. Edwards-Curve Digital Signature Algorithm (EdDSA) This document describes elliptic curve signature scheme Edwards-curve Digital Signature Algorithm (EdDSA). The algorithm is instantiated with recommended parameters for the edwards25519 and edwards448 curves. An example implementation and test vectors are provided. Ambiguity of Uppercase vs Lowercase in RFC 2119 Key Words RFC 2119 specifies common key words that may be used in protocol specifications. This document aims to reduce the ambiguity by clarifying that only UPPERCASE usage of the key words have the defined special meanings. CBOR Web Token (CWT) CBOR Web Token (CWT) is a compact means of representing claims to be transferred between two parties. The claims in a CWT are encoded in the Concise Binary Object Representation (CBOR), and CBOR Object Signing and Encryption (COSE) is used for added application-layer security protection. A claim is a piece of information asserted about a subject and is represented as a name/value pair consisting of a claim name and a claim value. CWT is derived from JSON Web Token (JWT) but uses CBOR rather than JSON. Concise Binary Object Representation (CBOR) The Concise Binary Object Representation (CBOR) is a data format whose design goals include the possibility of extremely small code size, fairly small message size, and extensibility without the need for version negotiation. These design goals make it different from earlier binary serializations such as ASN.1 and MessagePack. This document obsoletes RFC 7049, providing editorial improvements, new details, and errata fixes while keeping full compatibility with the interchange format of RFC 7049. It does not create a new version of the format. CBOR Object Signing and Encryption (COSE): Structures and Process Concise Binary Object Representation (CBOR) is a data format designed for small code size and small message size. There is a need to be able to define basic security services for this data format. This document defines the CBOR Object Signing and Encryption (COSE) protocol. This specification describes how to create and process signatures, message authentication codes, and encryption using CBOR for serialization. This specification additionally describes how to represent cryptographic keys using CBOR. This document, along with RFC 9053, obsoletes RFC 8152. CBOR Object Signing and Encryption (COSE): Initial Algorithms Concise Binary Object Representation (CBOR) is a data format designed for small code size and small message size. There is a need to be able to define basic security services for this data format. This document defines a set of algorithms that can be used with the CBOR Object Signing and Encryption (COSE) protocol (RFC 9052). This document, along with RFC 9052, obsoletes RFC 8152. Fully-Specified Algorithms for JSON Object Signing and Encryption (JOSE) and CBOR Object Signing and Encryption (COSE) This specification refers to cryptographic algorithm identifiers that fully specify the cryptographic operations to be performed, including any curve, key derivation function (KDF), and hash functions, as being "fully specified". It refers to cryptographic algorithm identifiers that require additional information beyond the algorithm identifier to determine the cryptographic operations to be performed as being "polymorphic". This specification creates fully-specified algorithm identifiers for registered JSON Object Signing and Encryption (JOSE) and CBOR Object Signing and Encryption (COSE) polymorphic algorithm identifiers, enabling applications to use only fully-specified algorithm identifiers. It deprecates those polymorphic algorithm identifiers. This specification updates RFCs 7518, 8037, and 9053. It deprecates polymorphic algorithms defined by RFCs 8037 and 9053 and provides fully-specified replacements for them. It adds to the instructions to designated experts in RFCs 7518 and 9053. CBOR Object Signing and Encryption (COSE) Receipts CBOR Object Signing and Encryption (COSE) Receipts prove properties of a Verifiable Data Structure (VDS) to a verifier. VDSs and associated Proof Types enable security properties, such as minimal disclosure, transparency, and non-equivocation. Transparency helps maintain trust over time and has been applied to certificates, end-to-end encrypted messaging systems, and supply chain security. This specification enables concise transparency-oriented systems by building on Concise Binary Object Representation (CBOR) and COSE. The extensibility of the approach is demonstrated by providing CBOR encodings for Merkle inclusion and consistency proofs. An Architecture for Trustworthy and Transparent Digital Supply Chains Traceability in supply chains is a growing security concern. While Verifiable Data Structures (VDSs) have addressed specific issues, such as equivocation over digital certificates, they lack a universal architecture for all supply chains. This document defines such an architecture for single-issuer signed statement transparency. It ensures extensibility and interoperability between different transparency services as well as compliance with various auditing procedures and regulatory requirements. Informative References Improving Awareness of Running Code: The Implementation Status Section This document describes a simple process that allows authors of Internet-Drafts to record the status of known implementations by including an Implementation Status section. This will allow reviewers and working groups to assign due consideration to documents that have the benefit of running code, which may serve as evidence of valuable experimentation and feedback that have made the implemented protocols more mature. This process is not mandatory. Authors of Internet-Drafts are encouraged to consider using the process for their documents, and working groups are invited to think about applying the process to all of their protocol specifications. This document obsoletes RFC 6982, advancing it to a Best Current Practice. HTTP Semantics The Hypertext Transfer Protocol (HTTP) is a stateless application-level protocol for distributed, collaborative, hypertext information systems. This document describes the overall architecture of HTTP, establishes common terminology, and defines aspects of the protocol that are shared by all versions. In this definition are core protocol elements, extensibility mechanisms, and the "http" and "https" Uniform Resource Identifier (URI) schemes. This document updates RFC 3864 and obsoletes RFCs 2818, 7231, 7232, 7233, 7235, 7538, 7615, 7694, and portions of 7230. HTTP Message Signatures This document describes a mechanism for creating, encoding, and verifying digital signatures or message authentication codes over components of an HTTP message. This mechanism supports use cases where the full HTTP message may not be known to the signer and where the message may be transformed (e.g., by intermediaries) before reaching the verifier. This document also describes a means for requesting that a signature be applied to a subsequent HTTP message in an ongoing HTTP exchange. Agent Transaction Protocol (ATP) x402: An Open Standard for Internet-Native Payments x402 Foundation Agent Payments Protocol (AP2) Google Agentic Commerce Triple Entry Accounting Summa de arithmetica, geometria, proportioni et proportionalita x402 STARK Receipt Format Extension Vauban Research Outcome Binding for Authorized Actions and Independently Observed Effects Compliance Profile of Signed Action Receipts for AI Agents Signed Decision Receipts for Machine-to-Machine Access Control Categorical Compliance Screening Receipt Format for Agentic-Payment Flows Cedulon Re-Attestation: Carrying Spend Evidence Across Algorithm Retirement Cedulon Streaming Reconciliation: Continuous Completeness for Agent Spend

Acknowledgments Vernon Wharff set out the defect -02 repairs: that the object carrying the T11 guarantee was neither profiled for registration nor read during verification, and that the equivocation requirement could not fire against a presented chain. He also asked the question that decided the shape of that repair, namely whether a recorded checkpoint absent from the chain deserves its own identifier or belongs under window coverage. Iman Schrock confirmed the finding independently and drew its boundary, keeping it separate from the extract-binding work already closed in -01. Iman Schrock raised the first of this revision's two subjects, against the posted -02: whether the profile should accept a pinned witness key and report an absent or mismatched pin explicitly. It should, and the same question turned out to be unanswered for three further objects. Iman Schrock and Pablo Play ran the -00 implementation against the pinned commit and reported the defects that produced -01. Iman Schrock found the two extract-binding defects, proposed the repair -01 adopts, later reran the posted -01 from a clean clone against its own pinned commit, and is also the author of , cited here as adjacent work. He is the reader whose independent implementation of the Signed Statement identity is noted in , and he asked for it to be kept separate from any cross-implementation claim about Cedulon; that separation is his and is recorded here as he stated it. Pablo Play found that a repeated reference hid the unaccounted amount, filed a written reproduction, and re-ran that reproduction against the pinned commit to confirm the figures quoted from it. He later took up a standing invitation to break the implementation and ran the suite on a platform its author had not, which is how the three defects behind 0.3.1 were found and how a fourth came to light while they were being repaired. Nicholas Templeman ran the suite from a clean clone and reported his figures. He also corrected two claims in a row written about that run: the install it named was not the strict from-lockfile form, and his platform was the same operating system family as the earlier ones, so the run corroborates the numbers and adds no cross-environment evidence. He classified his own run honestly as a repetition of the author's checks rather than an independent implementation. Walter Hawkins did not run it; he read the reported figures and pressed for the run to be stated precisely enough to be repeatable, which is why the conditions and not only the totals appear in . None of them reviewed this text, and any error in it is the author's. Field survey notes and the informative threat-model narrative in the companion repository helped shape the requirement identifiers used here. Those identifiers are defined in .

Appendix A. Test Vectors These vectors use RFC 8032 Ed25519 secret scalar #1 (fixture only; never a production key). Hex is lowercase. They MUST match the locked tests in the companion implementation. Receipt COSE_Sign1: Claims: payer=payer-1, payee=payee-1, amount=1, currency=USD, policyHash=aa, manifestHash=null, noManifest=true, x402PaymentRef=null, timestampMs=1700000000000, nonce=n100000000000000, prevReceiptHash=null, outcome=aborted. COSE_Sign1 hex (whitespace ignored; identical to the locked test): Manifest COSE_Sign1: Body: description=fixture-goods, amount=1, currency=USD, acceptanceCriteriaHash=00, cancelCondition=none, expiresAtMs=1700000000000, ap2MandateHash=null. COSE_Sign1 hex (whitespace ignored; identical to the locked test):