Wilder Robotics

rob@wilder-robotics.com

Security SCITT SCITT COSE receipts transparency physical AI robotics trusted execution environment attestation regulated site This document defines a SCITT profile for Physical-Site Engagement Receipts (PSER): tamper-evident, signed, offline-verifiable records that describe an autonomous or human-directed physical engagement at a specific real-world site governed by a defined operating envelope. Each receipt is a SCITT Signed Statement as defined by the SCITT architecture, encoded as a COSE Single Signer message, carrying a JCS-canonicalized JSON payload with a five-artifact vocabulary describing (1) the Site, (2) the Operator and Actor, (3) the Engagement Window and Envelope, (4) the Attestation Evidence from a Trusted Execution Environment (TEE), and (5) the Adapter Write-In recording that the receipt was posted into an out-of-band operations layer. A Physical-Site Engagement Receipt is registerable in any conforming SCITT Transparency Service, obtaining a Receipt that proves the Statement's inclusion in that Service's verifiable data structure. Registration does not establish that the Issuer registered every receipt it issued. Profile 0.7 additionally binds separately retained, attributed content facts through a salted commitment and selected disclosures. Integrity and permitted attribution do not establish the facts' truth or the asserting parties' identity. This profile deliberately makes a NARROW, checkable claim -- "this is a tamper-evident, signature-verifiable record that a specific engagement occurred at a specific site under a specific envelope, and its evidence was sealed by a specific TEE" -- and explicitly does NOT claim that the engagement was safe, correct, or wise, that the site conditions were as described, or that any downstream operational outcome followed. Compliance verdicts derived from the receipt (SLA credit, insurance underwriting, regulatory audit) are the responsibility of the relying party and its policies, not of this profile. The profile is designed around a three-party trust model in which no single party can unilaterally forge or repudiate a receipt: the Site Owner controls physical access to the TEE hardware and keeps it running (they can unplug the box, and cannot forge what it signs); the TEE silicon vendor attests the key material inside the TEE through its hardware root of trust (silicon vouches for the key); and the Issuer writes the vocabulary, registers Signed Statements with a Transparency Service, and posts the resulting receipt into the site's operations layer via a WRITE_ONLY adapter. This separation is normative in this profile: implementations MUST NOT collapse these three roles into a single custodian, and relying parties MUST NOT trust a receipt that lacks any one of them.

Introduction Autonomous mobile robots, semi-autonomous physical equipment, and human-directed physical work crews increasingly operate at regulated real-world sites -- warehouses, common-interest communities, industrial facilities, healthcare campuses, and public infrastructure. Relying parties -- site owners, insurers, regulators, dispatchers, and downstream operations platforms -- need portable, verifiable evidence of what physically happened at a site, distinct from the digital-artifact supply-chain evidence addressed by and distinct from the per-action AI-agent evidence addressed by and . This profile fills that gap by defining the SCITT Statement content for one physical-site engagement: a bounded interval during which a specific actor operates at a specific site under a stated envelope, with the evidence sealed inside a TEE and the receipt subsequently written into whatever operations layer the site already uses (property-management system, maintenance ticketing, insurance underwriting API, regulatory portal). The profile's defensibility, and its value to relying parties, comes from combining four elements that no single vendor category currently ships together:

• Site-hosted TEE trust anchor. The signing key is bound to hardware physically located at the site under the site owner's control. Cloud-hosted transparency services can issue strong receipts, but the signing authority lives inside the cloud provider's environment; this profile REQUIRES that the authority live on the Site Owner's premises, attested by the TEE silicon vendor, and neither extractable by the Site Owner nor by the Issuer. This statement describes direct-witness mode. In delegated-witness mode () the key that produces the COSE signature is the Issuer's own and need not be site-resident; what remains site-resident is the TEE that issues the delegation credential, and a relying party evaluating such a receipt obtains a weaker property than the one described here. A Verifier MUST determine which mode applies from attestation.bindingMode in the receipt before relying on the non-extractability property, and MUST NOT assume direct-witness mode. A receipt carrying no attestation.bindingMode is rejected on version validation under and no mode is inferred for it.
• Physical-work evidence vocabulary. The five-artifact schema (Site, Actor, Engagement, Attestation, Adapter Write-In) binds the receipt to what physically happened, not merely to a software event. This vocabulary is defined in and is stricter than a general-purpose SCITT Statement.
• WRITE_ONLY adapter into existing operations layers. Verified evidence is posted into the systems the buyer already uses -- property-management, maintenance, warehouse-management, claims, and asset-management platforms -- as recorded by the adapter field in . This profile explicitly does NOT define a new operations dashboard; it defines how receipts enter the operations layers a site already runs.
• Transparency-service registration. Neither the Issuer's chain nor the TEE establishes that a presented history is complete, or that it is the only history. A withheld suffix is internally consistent at every link, and a TEE establishes that it wrote the state it attests, not that that state is the most recent. Registration in a SCITT Transparency Service supplies the external reference against which relying parties and auditors can test those questions. A TEE on customer premises without external witnessing is therefore insufficient; SCITT registration is REQUIRED by this profile ().

Physical-Site Engagement Receipts are complementary to, and compose with, existing SCITT-AI drafts. An AI agent that dispatches a physical robot MAY emit an Agent Action Capsule per describing the dispatch decision, and the physical engagement that follows MAY be recorded as one or more Physical-Site Engagement Receipts under this profile. An application can correlate these records using explicitly identified references and its selected policy. Under profile 0.7, CWT sub remains exactly equal to site.id; shared subject text is not evidence that records concern the same engagement or establish causation. See . The architecture requirements stated here are not evidence that a particular implementation satisfies them. Revision -06 retains the 0.7 wire format and adds informative composition guidance in . A signed record, a verified registration, native hardware appraisal, and an application decision remain distinct results.

Requirements Notation 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.

Non-goals This revision does not:

• Attest that the engagement was safe, correct, effective, or compliant with any specific regulation.
• Attest that the site conditions were as recorded.
• Attest that no unrecorded engagement occurred outside the instrumented boundary.
• Specify a deterministic offline replay of a physical engagement decision. Reproducing a recipient's checks over retained bytes is a different operation and is described for the local implementation in .
• Define the operations-layer schemas the Adapter Write-In targets.
• Define billing, SLA-credit, or insurance-pricing rules that a relying party may derive from a stream of receipts.
• Attest anything about the internal state, intent, or decision process of a human participant in an engagement, or about signals conveyed by a direct neural or brain-computer interface. This profile records that a bounded physical engagement occurred at a Site and identifies the parties that can attest to it. A direct neural interface is not an engagement performed by an Actor at a Site under , and this document defines no member, no value, and no extension point for one.

These non-goals are NORMATIVE: implementations and relying parties MUST NOT imply the stronger claims from a receipt.

Terminology This document uses the terms defined in (Signed Statement, Statement, Issuer, Subject, Transparency Service, Registration Policy, Receipt) and (Verifiable Data Structure, Verifiable Data Structure Proof). In addition:

Site:

The bounded real-world location at which the engagement occurred, identified by a stable Site Identifier under the Issuer's registration authority. The Site is the physical analog of a SCITT Subject.

Site Envelope:

The operating constraints in force at the Site during the engagement -- permitted actor classes, permitted engagement types, geospatial bounds, temporal bounds, and referenced site-rule documents. The Site Envelope is identified by a stable envelope identifier and a content digest.

Site Owner:

The party that controls physical access to the TEE hardware producing receipts for a Site, and that is responsible for that hardware's continued operation there. The Site Owner is defined by those two capabilities and not by title, by legal ownership of the premises, or by any contractual label: the party holding them may or may not be the party named on the deed. The Site Owner is one of the three parties REQUIRED to participate in every receipt under . The Site Owner has no capability to author, alter, or suppress the content of a receipt, and none to extract the signing key material.

Actor:

The physical entity that performed the engagement -- an autonomous robot, a semi-autonomous asset, a human operator, or a human-led crew -- identified by a stable actor identifier under the Issuer's registration authority.

Operator:

The organization or individual responsible for the Actor during the engagement, distinct from the Issuer of the receipt when a third-party witness signs.

Engagement:

A bounded interval, delimited by an Engagement Window, during which the Actor performed physical work at the Site under the Site Envelope.

Engagement Window:

The time interval [start, end] of the Engagement, expressed in RFC 3339 UTC, with the same clock basis as the TEE-sealed evidence.

Attestation Evidence:

The output of a TEE that observed the Actor and the Engagement, including a platform attestation, a measured-boot chain, and a digest over the sealed evidence bundle. The bundle itself is opaque to the Transparency Service.

Adapter Write-In:

The record that the Signed Statement (or a reference to it) was posted into an out-of-band operations layer, together with the operation-layer system identifier, endpoint identifier, and a post-time digest of the operations-layer acknowledgement. The Adapter Write-In is what makes the receipt useful to the site's existing workflow without requiring the operations layer to be modified.

Physical-Site Engagement Receipt (PSER):

A SCITT Signed Statement under this profile, carrying a canonical JSON payload conforming to , with the profile identifier wilder.pser/0.7 and a SCITT Receipt attached as defined in .

Chain-Verifier:

A relying party, or a party acting on a relying party's behalf, that is presented with two or more Physical-Site Engagement Receipts as one contiguous chain and evaluates the chain-level properties defined in . Chain-Verifier is a role, not a distinct principal: any verifier MAY act as a Chain-Verifier, and the obligations this profile places on a Chain-Verifier apply only to a presentation of two or more receipts. A verifier presented with a single receipt incurs none of them.

Profile identifier and media types The profile identifier for this document is wilder.pser/0.7 and MUST appear as the top-level payload spec value. The protected COSE content type MUST be application/pser+json; profile=wilder.pser/0.7, and it MUST agree with the payload's declared profile. Profile 0.7, introduced in revision -05, adds the required null-or-root engagement.contentDigest and makes the Signed Statement's protected CWT sub a text string exactly equal to site.id. It retains the 0.6 recorded-time containment and DIRECT_WITNESS identifier-consistency rules. Revision -06 retains those requirements without adding a profile version or authenticated key association. Statements declaring wilder.pser/0.5 or wilder.pser/0.6 retain their earlier profile requirements. An implementation supporting legacy material MUST NOT silently reinterpret it as 0.7, insert a missing content member, rewrite its CWT types, or normalize its signed bytes. A 0.7 recipient MUST report unknown or unsupported profiles as unsupported, not substitute a known version with different rules. The legacy implementation's byte-string sub form is preserved as compatibility behavior, not represented as conforming to the CWT text claim definition in or the SCITT architecture in . Existing signatures are not repaired by changing the encoded claim type. The application/scitt-statement+cose and application/scitt-receipt+cose media types from apply unchanged to the outer objects. The retained content is not inline public geometry or a new SCITT log format.

Receipt structure A Physical-Site Engagement Receipt is a SCITT Signed Statement per Section 6, encoded as a COSE_Sign1 per . The payload is a JSON object serialized with JCS and carried as the COSE_Sign1 payload. The following is a complete example instance. It is not a schema: every value is literal, the whole object parses as JSON, and the chain.hash value is the digest this profile specifies over the rest of the object. Normative member definitions are in ; the member definitions govern over an illustrative example. This figure is illustrative. Its digests are placeholders, the identifiers are synthetic, and the teeClass value is one conforming registry entry chosen so the example round-trips. The chain.hash value is computed over the canonicalized payload with chain.hash absent. This profile does not prefer, presume, or depend on any particular confidential-compute environment, and no value in this figure should be read as a statement about deployed hardware.

Physical-Site Engagement Receipt payload { "actor": { "class": "AUTONOMOUS", "id": "actor:robot-alpha-01", "operator": "operator:wilder-robotics" }, "adapter": { "ackDigest": "sha256:4444444444444444444444444444444444444444444444444444444444444444", "ackProvenance": "THIRD_PARTY", "endpoint": "endpoint:res-001", "mode": "WRITE_ONLY", "postedAt": "2026-10-15T14:00:05Z", "system": "example.ticketing" }, "attestation": { "bindingMode": "DIRECT_WITNESS", "measuredBoot": { "chain": "sha256:98a6efd412bb768ea7f090e8228401c11bc72a7caae44170395445c097d5ffa1", "components": [ { "digest": "sha256:2222222222222222222222222222222222222222222222222222222222222222", "name": "bl1" } ] }, "platformEvidence": { "digest": "sha256:cccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc", "encoding": "opaque/1" }, "sealedEvidence": { "digest": "sha256:3333333333333333333333333333333333333333333333333333333333333333", "encoding": "opaque/1", "sizeBytes": 4096 }, "teeClass": "arm.cca", "validity": { "notAfter": "2026-10-15T15:00:00Z", "notBefore": "2026-10-15T13:00:00Z" }, "witnessKey": "key:tee:res-001-witness-01" }, "chain": { "hash": "sha256:d9818729036b31e66078dade02bc83dc69cb5935ac226978d2fede70bde4a1bb", "prevHash": null, "seq": 0 }, "engagement": { "contentDigest": null, "envelopeConformance": "WITHIN", "evidenceDigest": "sha256:1111111111111111111111111111111111111111111111111111111111111111", "id": "eng:res-001:20261015-140000", "outcomeClass": "COMPLETED", "type": "patrol", "window": { "end": "2026-10-15T14:00:00Z", "start": "2026-10-15T13:30:00Z" } }, "id": "uuid:00000000-0000-4000-8000-000000000001", "issuerAffiliation": "NOT_DISCLOSED", "site": { "class": "residential", "envelope": { "digest": "sha256:0000000000000000000000000000000000000000000000000000000000000000", "geobounds": null, "id": "env:res-001:2026-Q4", "temporal": { "ends": null, "starts": "2026-10-01T00:00:00Z" } }, "id": "site:res-001" }, "spec": "wilder.pser/0.7", "ts": "2026-10-15T14:00:00Z" }

Field semantics

spec (REQUIRED, string) MUST be wilder.pser/0.7 for receipts conforming to this document. An unsupported value is outside the recipient's implemented profile set and MUST be reported as unsupported. Implementations supporting 0.5 or 0.6 evaluate those statements under their declared versions, without importing the new content or subject-type requirements.

id (REQUIRED, string) A globally unique identifier for the receipt, assigned by the Issuer. RECOMMENDED form is a URN or a uuid: prefix. id MUST NOT be reused within an Issuer.

ts (REQUIRED, string) RFC 3339 UTC timestamp at which the Issuer sealed the receipt. This is the receipt-issuance time; it MAY differ from engagement.window.end. For a receipt declaring wilder.pser/0.6 or wilder.pser/0.7, the Verifier MUST check notBefore <= ts <= notAfter using parsed instants and inclusive endpoints. The comparison checks whether the asserted receipt-issuance timestamp lies within the asserted attestation-validity interval. It does not independently establish the actual time of issuance or whether the engagement occurred within that interval. A timestamp at either endpoint passes this check only; it does not establish overall receipt conformance. This containment requirement does not apply to a receipt declaring wilder.pser/0.5. An implementation supporting these profile versions evaluates each under its declared version's requirements. A verifier-selected clock-skew tolerance MUST NOT alter the profile's containment interval.

issuerAffiliation (REQUIRED, string) States whether the Issuer and the Site Owner are affiliated principals. The admissible values are exactly:

• AFFILIATED: the Issuer and the Site Owner are the same principal, or are principals under common control, or one controls the other.
• INDEPENDENT: the Issuer and the Site Owner are principals under neither common control nor the control of one by the other.
• NOT_DISCLOSED: the relationship is not stated in the receipt.

The member is REQUIRED because the alternative is worse. An absent member would itself have to be assigned a meaning, and every available meaning is wrong: read as INDEPENDENT it manufactures a disclosure nobody made, and read as AFFILIATED it accuses an Issuer of a relationship it may not have. Requiring the member makes NOT_DISCLOSED a stated position rather than an inference drawn from silence. The value states the Issuer's own claim about itself. This profile does not define a mechanism by which a Verifier establishes the claim to be true, and a Verifier MUST NOT report a verified receipt as evidence that the stated relationship holds. What verification establishes is that the claim was made, by the Issuer, inside a receipt bound by the signature and the chain, and therefore that it cannot later be revised without the revision being visible. That is a narrower property than truth and it is the property this member carries. A Verifier MUST NOT read NOT_DISCLOSED as INDEPENDENT. Silence about a relationship is not a denial of one, and a relying party told otherwise has been supplied a claim no principal authored. This is the one collapse the member exists to prevent, and it is the direction that overstates the receipt. A Verifier MUST NOT read a value outside the admissible set as AFFILIATED and MUST NOT normalize it to NOT_DISCLOSED. A Verifier that encounters an unrecognized value MUST preserve the value as received and MUST surface it to the relying party as unrecognized, distinct from all three admissible values. This revision does not require a Verifier to reject a receipt on that basis, because a value it does not recognize may be defined by a later revision; it requires that the Verifier never silently resolve the ambiguity. The value set is closed in this revision and is not registry-governed. A later revision that adds a value does so additively, without redefining an existing one. The relationship this member describes is a standing one between two principals rather than a property of a single engagement, but standing relationships change. Receipts presented as one chain may therefore disagree about it without either receipt being defective. The chain-level obligation, which is to surface such a change rather than to reject the presentation or to resolve it in favour of either value, is stated with the other Chain-Verifier obligations under chain. This member is distinct from, and does not substitute for, the affiliation disclosure required of an Issuer that registers with a Transparency Service it operates or that is operated by an affiliated principal (), which is an Issuer-published fact resolved under rather than a member of this payload. That obligation concerns the relationship between the Issuer and the Transparency Service; this member concerns the relationship between the Issuer and the Site Owner. An Issuer may be independent of the Site Owner and still operate its own Transparency Service, or be affiliated with the Site Owner and register with an unaffiliated one. Neither value can be inferred from the other.

site (REQUIRED, object) Identifies the physical location.

• site.id (REQUIRED, string): stable site identifier under the Issuer's registration authority. This is the physical analog of a Subject and MUST be used as the text value of the CWT sub claim for 0.7 in the protected header (see ).
• site.class (REQUIRED, string): coarse site classification. Registry- governed; see .
• site.envelope.id (REQUIRED, string): stable identifier of the operating envelope in force during the engagement.
• site.envelope.digest (REQUIRED, string): JSON-DIGEST (SHA-256 of the JCS serialization) of the full envelope document. The full document MUST NOT appear in the public receipt; it is bound by digest only.
• site.envelope.geobounds (OPTIONAL, string): opaque reference to geospatial bounds. Any geospatial detail beyond the reference is bound by the envelope digest, not published.
• site.envelope.temporal (OPTIONAL, object): temporal window during which this envelope was in force. null values indicate "open-ended in that direction."

actor (REQUIRED, object)

• actor.id (REQUIRED, string): stable identifier of the physical actor.
• actor.class (REQUIRED, string): one of AUTONOMOUS, SEMI_AUTONOMOUS, HUMAN, CREW.
• actor.operator (REQUIRED, string): stable identifier of the responsible operator organization or individual.

engagement (REQUIRED, object)

• engagement.id (REQUIRED, string): stable identifier of the engagement.
• engagement.window.start and engagement.window.end (REQUIRED, string): RFC 3339 UTC bounds of the engagement. end MUST be >= start. Both MUST share a clock basis with attestation.sealedEvidence (see ).
• engagement.type (REQUIRED, string): coarse engagement classification (e.g. patrol, service, inspection, delivery, installation, maintenance, presence). Registry-governed; see .
• engagement.outcomeClass (REQUIRED, string): one of COMPLETED, ABORTED, REFUSED, ERRORED, OBSERVED_ONLY. OBSERVED_ONLY records that the Issuer witnessed the actor at the site but did not participate in dispatch.
• engagement.envelopeConformance (REQUIRED, string): one of WITHIN, EXCEEDED_TEMPORAL, EXCEEDED_GEO, EXCEEDED_ACTOR, UNKNOWN. The Issuer MUST NOT claim WITHIN unless it evaluated conformance against the envelope digest.
• engagement.evidenceDigest (REQUIRED, string): JSON-DIGEST of the engagement's internal evidence structure. The internal structure is opaque to this profile and MUST NOT appear in the receipt.
• engagement.contentDigest (REQUIRED for 0.7): JSON null or the sha256: root of the retained-content construction in . It is distinct from evidenceDigest and the whole-envelope digest. A missing member is invalid for 0.7. A payload declaring 0.5 or 0.6 that contains this member is invalid under that version and MUST be rejected, including when its value is null. This does not authorize modifying an existing signed statement.

attestation (REQUIRED, object) Binds the receipt to the TEE that observed the engagement. This is the mechanism that distinguishes a Physical-Site Engagement Receipt from a bare signed timestamp: the sealed evidence attests that the Issuer observed the engagement from inside a hardware-rooted, remotely attestable environment. Verification of the Signed Statement and its attached Receipt alone does not establish hardware provenance. That property additionally depends on obtaining and validating the platform attestation evidence and its binding to the relevant key under the relying party's trust policy.

• attestation.teeClass (REQUIRED, string): TEE class identifier. Registry-governed; see . The TEE Class registry is REQUESTED by this document and has NOT yet been allocated by IANA. Until allocation, the admissible values are exactly the initial values listed in . A Verifier MUST reject a teeClass value outside that set. These descriptions identify the environments named by the existing values. They do not, by themselves, define a complete evidence-format binding or establish implementation support. Evidence encoding, versioning, trust inputs, and appraisal requirements are addressed by the applicable evidence specification and the profile's attestation model (attestation.platformEvidence, ).
• attestation.platformEvidence (REQUIRED, object): reference to the platform-native attestation document, in a format defined by the TEE class. The document itself MAY be conveyed by reference (URI + digest) or inline; when conveyed inline it SHOULD be in the unprotected header of the enclosing Signed Statement, not in the payload.
• attestation.platformEvidence.digest (REQUIRED, string): digest of the platform-native attestation document.
• attestation.platformEvidence.encoding (REQUIRED, string): opaque encoding label for that document. The set of labels a given TEE class admits is defined by that TEE class.
• attestation.measuredBoot (REQUIRED, object): the measured-boot state of the environment that produced the receipt.
• attestation.measuredBoot.chain (REQUIRED, string): JSON-DIGEST of the measured-boot chain.
• attestation.measuredBoot.components (REQUIRED, array): the measurements the chain digest commits to, in boot order. Each element is an object with a name (REQUIRED, string) naming the measured component and a digest (REQUIRED, string) carrying its measurement. Verifiers MUST NOT infer any meaning from name beyond identification.
• attestation.sealedEvidence.digest (REQUIRED, string): digest of the sealed evidence bundle.
• attestation.sealedEvidence.sizeBytes (REQUIRED, int): size of the sealed bundle in bytes. Included to enable bounded-storage verifiers to reject bundles they cannot process.
• attestation.sealedEvidence.encoding (REQUIRED, string): opaque encoding label. Registry-governed; see .
• attestation.witnessKey (REQUIRED, string): key identifier of the TEE signing key. This MAY differ from the Issuer's iss when the TEE operates as a delegated witness. Under the wilder.pser/0.6 and wilder.pser/0.7 DIRECT_WITNESS naming convention, attestation.witnessKey and the protected CWT iss value MUST be textually equal. A successful comparison establishes this naming convention only. It does not establish that the signature-verification key is authentically associated with either identifier or that genuine TEE hardware produced the signature. This convention does not require a global one-to-one mapping between all issuer identifiers and keys, prohibit key rotation, or redefine issuer identity as key material. The actual DIRECT_WITNESS relationship still concerns the key that produced the signature and the TEE signing key. Authenticated key association and hardware provenance require their respective evidence and trust mechanisms; they are not supplied by label equality. This naming restriction MUST NOT be silently applied as a changed conformance rule for wilder.pser/0.5.
• attestation.bindingMode (REQUIRED, string): the attestation-binding mode under which this receipt was produced, as defined in . The admissible values are exactly:
  • DIRECT_WITNESS: the key that produced the COSE signature is the TEE signing key. attestation.bindingMode is DIRECT_WITNESS only where attestation.witnessKey and iss denote the same key.
  • DELEGATED_WITNESS: the key that produced the COSE signature is the Issuer's own, and a TEE-issued delegation credential authorizes it.
  A Verifier MUST reject a bindingMode value outside that set, and MUST reject a receipt asserting DIRECT_WITNESS in which attestation.witnessKey and iss denote different keys. The value is closed in this revision and is not registry-governed. This member is REQUIRED, and carries in the receipt a fact that -02 required a Verifier to obtain from the Issuer out of band. The mode was fixed at the moment the receipt was signed and was known to the signer; obtaining it from a separately published document made a per-receipt fact depend on a document that describes an Issuer rather than a receipt, and made the weaker of the two properties in unavailable from the presented bytes. An absent value is not assigned a meaning, because assigning one reintroduces the assumption this member exists to prevent.
• attestation.validity (REQUIRED, object): the interval over which the attestation of the producing environment is asserted to hold.
• attestation.validity.notBefore (REQUIRED, string): RFC 3339 UTC timestamp at which the attestation becomes valid.
• attestation.validity.notAfter (REQUIRED, string): RFC 3339 UTC timestamp after which the attestation is no longer valid. notAfter MUST be strictly later than notBefore; a Verifier MUST reject a receipt whose notAfter is equal to or precedes its notBefore. A zero-length interval asserts validity for an instant of zero duration and has no legitimate producer. For a receipt declaring wilder.pser/0.6 or wilder.pser/0.7, a Verifier MUST reject the receipt when the asserted ts is outside the asserted [notBefore, notAfter] interval, using parsed instants and inclusive endpoints. This checks consistency of the recorded instants, not independently established real-world issuance time. The published wilder.pser/0.5 profile did not require timestamp containment; that historical requirement remains unchanged.

adapter (REQUIRED, object) Records that the receipt (or a reference to it) was written into an out-of-band operations layer. This is the profile's core insight: a receipt that no operations system can see is not useful, and modifying the operations system to consume receipts natively is out of scope for most regulated sites. The Adapter Write-In makes the receipt observably present in the site's existing workflow.

• adapter.system (REQUIRED, string): operations-layer system identifier (e.g. a property-management system, maintenance ticketing platform, regulatory portal, insurance underwriting API). Registry-governed; see .
• adapter.endpoint (REQUIRED, string): opaque endpoint identifier within the system. Its interpretation is defined by the target system, not by this profile.
• adapter.postedAt (REQUIRED, string): RFC 3339 UTC timestamp at which the write-in was posted.
• adapter.ackDigest (REQUIRED, string): JSON-DIGEST of the operations- layer's acknowledgement response. If the operations layer returns no structured acknowledgement, the digest is taken over an Issuer-defined minimal ack object. That object is bound by the receipt's Merkle inclusion and is not published, so no Verifier can obtain it and none is required to. This is not a resolution obligation under : a Verifier checks that adapter.ackProvenance is ISSUER_ASSERTED and treats the acknowledged content as authored by the Issuer. -02 located the object's schema in an Issuer-published document, which stated an obligation against a document defined to be unfetchable.
• adapter.ackProvenance (REQUIRED, string): identifies which party authored the acknowledgement that adapter.ackDigest commits to. adapter.ackDigest alone cannot carry this: the digest of an acknowledgement authored by an independent operations layer and the digest of one authored by the Issuer itself are indistinguishable to a Verifier, and the two have materially different evidentiary weight. The admissible values are exactly:
  • THIRD_PARTY: the acknowledgement was returned by the operations layer named in adapter.system, which is a principal distinct from the Issuer.
  • ISSUER_ASSERTED: the operations layer returned no structured acknowledgement, and the digest is taken over the Issuer-defined minimal ack object described under adapter.ackDigest. The Issuer is the author of the acknowledged content.
  • NONE: no acknowledgement was obtained from any party.
  A Verifier MUST NOT read a value outside that set as THIRD_PARTY, and MUST NOT normalize it to NONE. Doing either reintroduces the collapse this member exists to prevent, in the direction that overstates the receipt. A Verifier that encounters an unrecognized value MUST preserve the value as received and MUST surface it to the relying party as unrecognized, distinct from all three admissible values. This revision does not require a Verifier to reject a receipt on that basis, because a value it does not recognize may be defined by a later revision; it requires that the Verifier never silently resolve the ambiguity in the receipt's favour. The value set is closed in this revision and is not registry-governed. A later revision that adds a value does so additively, without redefining an existing one.
• adapter.mode (REQUIRED, string): MUST be WRITE_ONLY in this revision. Read-in modes are explicitly out of scope; see .

chain (REQUIRED, object) Hash-chains successive receipts by the same Issuer so that a verifier can detect broken hash links, sequence discontinuities, and modification, substitution or reordering among the receipts presented as one contiguous chain. The chain does NOT establish that its last presented receipt is the Issuer's latest: a prover that withholds a suffix presents a prefix that is internally consistent at every link. See . The construction is defined normatively in this document. It follows the convention established in Section 5, which is cited for provenance only: no conformance requirement of this profile depends on that document.

• chain.seq (REQUIRED, int): non-negative sequence number within the Issuer's chain for the identified Subject. The first receipt in a chain MUST carry chain.seq 0.
• chain.prevHash (REQUIRED, string or null): the value of the immediately preceding receipt's chain.hash, or null for the first receipt. A receipt whose chain.seq is 0 MUST carry null; a receipt whose chain.seq is nonzero MUST carry the preceding receipt's chain.hash value. Note that this is a digest over the preceding receipt EXCLUDING its chain.hash member, per the definition of chain.hash below; it is not a digest over the preceding receipt as transmitted.
• chain.hash (REQUIRED, string): JSON-DIGEST of the receipt's canonical form with the chain.hash member absent. An Issuer computes this value over the complete receipt including chain.seq and chain.prevHash, then inserts it; a verifier recomputes it by removing the member before canonicalizing. chain.hash is never an input to its own computation.

A Chain-Verifier presented with two or more receipts as one contiguous chain MUST check, for each adjacent pair, that the later receipt's chain.seq is exactly one greater than the earlier receipt's, and that the later receipt's chain.prevHash equals the earlier receipt's chain.hash. A verifier that does not perform both checks MUST NOT report the presentation as a verified chain. These are chain-level obligations; an Issuer producing individual receipts is unaffected by them. A Chain-Verifier MUST additionally compare issuerAffiliation across each adjacent pair. Where two receipts presented as one chain carry different values, the Chain-Verifier MUST surface the change to the relying party, identified by the chain.seq of the receipt carrying the later value. A change in issuerAffiliation does not by itself invalidate the presentation, and a Chain-Verifier MUST NOT report the presentation as unverified on that basis alone. The two preceding checks are structural. chain.seq and chain.prevHash are wholly under the Issuer's control, so a violation of either admits no honest explanation. issuerAffiliation is not structural: it states a relationship between two principals in the world outside the receipt, and such relationships change. An Issuer independent of a Site Owner at one engagement may be acquired by that Site Owner before the next. Reporting that as an unverified chain would place an ordinary corporate event in the same category as tampering, and the only conforming response available to the Issuer would be to begin a new chain, which resets chain.seq and chain.prevHash and so severs the record either side of the change. That is the continuity the chain exists to carry. What a Chain-Verifier MUST NOT do is reduce the presentation to a single affiliation value. In particular it MUST NOT adopt the value carried by the latest receipt as the value of the chain. Adopting the later value would allow a chain to be relabelled after the fact by appending a single receipt, with nothing in the presentation showing that the label had previously said something else. Each reported value remains attached to the receipts that carry it. A conforming three-receipt chain, the sequence-gap and broken-link cases these checks are required to reject, and a two-receipt presentation whose members disagree about issuerAffiliation and which is required to verify with the change surfaced, are published as test data in the reference implementation repository. Implementers are advised to confirm that an honest complete chain verifies under their implementation before relying on any of these checks.

COSE header requirements The protected header of a Signed Statement under this profile MUST include the CWT Claims header parameter (label 15, ), carrying at least:

• iss (CWT claim label 1): a URI identifying the Issuer, encoded as a nonempty CBOR text string.
• sub (CWT claim label 2): a CBOR text string whose decoded UTF-8 value is exactly the payload's site.id.

For 0.7, a byte-string subject, a missing subject, or a different text value MUST be rejected by the implemented subject check; no byte-to-text coercion, case folding, or Unicode normalization is permitted. This makes the encoded claim follow , , and . The equality check does not authenticate the Site. The protected header content_type (label 3) MUST be application/pser+json; profile=wilder.pser/0.7. The Signed Statement's payload MUST be the JCS serialization of the JSON object defined in . Detached payloads are NOT PERMITTED under this revision.

Resolving Issuer-published facts Two obligations in this profile require a Verifier to obtain a fact the Issuer publishes rather than carries in the receipt: the delegation credential of , and the Transparency Service affiliation disclosure of . Both are disclosures about the Issuer. Neither is an identity claim, and neither is required to evaluate a receipt produced in direct-witness mode by an Issuer registering with an unaffiliated Transparency Service. This revision does not specify a serialization format for these facts, and does not define a document that carries them. -02 referred to an "Issuer's manifest" in four normative requirements without defining one, so a Verifier was four times required to read something the profile never described. Naming the obligations and their resolution behaviour, and leaving the encoding to a subsequent revision or companion document, is deliberate: a format fixed before any has been deployed is more likely to be repudiated by the next revision than refined by it. An Issuer-published fact is resolved as follows.

• The Issuer MUST make the fact retrievable at a stable identifier under its own control, and that identifier MUST be discoverable from iss.
• A Verifier MAY cache a resolved fact. A cached answer MUST NOT survive a change in the signing key it was resolved for; on such a change the Verifier MUST re-resolve.
• Where a fact does not resolve, whether because it is unreachable, absent, or unreadable, the fact is undetermined.

Undetermined is a third outcome, not a synonym for either answer. A Verifier MUST surface an undetermined fact as undetermined. It MUST NOT resolve an undetermined fact to whichever value favours the Issuer, and MUST NOT report that a fact was absent where it was never successfully retrieved: those two states are distinct and a relying party's policy may treat them differently. A Verifier MUST NOT reject a receipt solely because an Issuer-published fact is undetermined; whether an undetermined fact is disqualifying is a policy question for the relying party and is out of scope for this profile. What the profile requires is that the relying party be told. Conformance vectors accompanying this revision MUST include, for each Issuer-published fact, at least one case in which resolution fails, and the expected outcome of such a case MUST NOT be acceptance.

Attestation binding The attestation.witnessKey field carries the identity of the TEE signer. This profile permits two attestation-binding modes. The mode under which a receipt was produced MUST be carried in that receipt, in attestation.bindingMode (), and MAY additionally be recorded in the CWT Claims Set. It is not obtained from any Issuer-published document:

• Direct-witness mode: the signing key associated with the Issuer identified by iss is the TEE signing key. Under wilder.pser/0.6 and wilder.pser/0.7, attestation.witnessKey and the protected CWT iss value MUST be textually equal under the naming convention defined in this profile. That comparison establishes identifier consistency only; it does not establish authenticated association with the key used to verify the signature or genuine hardware provenance. The published wilder.pser/0.5 requirement to reject a DIRECT_WITNESS receipt whose identifiers denote different keys remains part of that earlier profile. The string comparison retained by 0.6 and 0.7 neither changes the earlier profile nor supplies the missing authenticated-key-association mechanism. The reference implementation's broader assurance limitation is recorded in .
• Delegated-witness mode: the Issuer's iss key is distinct from the TEE signer, and the TEE has issued a delegation credential authorizing the Issuer to sign this receipt on the TEE's behalf. The delegation credential is bound by the attestation.sealedEvidence.digest and is an Issuer-published fact resolved under . Where it does not resolve, the authorization of the signing key is undetermined and the Verifier proceeds as required by that section. Delegated signing does not itself establish evidence appraisal.

Evidence model and external appraisal This revision retains the existing attestation payload structure and introduces neither an attestation.quote member nor an attestationResult member. The payload carries references, digests, and attestation-related claims; measuredBoot also carries the listed component measurements. The existing allowance for the platform attestation document to be supplied by reference or inline in the enclosing Signed Statement's unprotected header is unchanged. A relying party may appraise obtained Evidence itself or use a separate Verifier channel. A digest commits to particular evidence bytes; it is not, by itself, a retrieval address, a retrieval mechanism, or evidence appraisal. Obtaining evidence whose digest matches a receipt is distinct from validating that evidence and its binding to the relevant key. An entity performs the RATS Verifier role only when it appraises Evidence using the applicable trust inputs and appraisal policy and produces an Attestation Result. A single entity may perform several roles, but their combination must be stated rather than inferred from the signing mode. Verification of the Signed Statement and attached Receipt alone does not establish hardware provenance. The choice of any additional in-receipt Attestation Result representation remains open. Its format, authentication, evidence/key binding, freshness, and failure behavior are not specified by this revision. Option D (retain evidence references and additionally support an optional Attestation Result) is a future design direction, not an added 0.7 payload member. These functional distinctions follow the RATS role definitions ().

Clock basis All timestamps in a Physical-Site Engagement Receipt MUST share a single clock basis: the clock the TEE observed at the time it sealed the evidence bundle. Implementations MUST NOT mix wall-clock timestamps with TEE-observed timestamps within a single receipt. Verifiers MUST derive elapsed-time computations from the receipt's own bytes, not from the verifier's local wall clock.

Retained content blocks and fact provenance This section defines the retained-content construction for wilder.pser/0.7. It does not modify a 0.5 or 0.6 statement. Those legacy profiles reject engagement.contentDigest, including an explicit null value. A content block is a collection of recorded assertions, not a second SCITT Statement, a log Receipt, a geometry document, or an assertion that all relevant observations were captured. A content producer prepares the retained block; the Issuer binds its root in engagement.contentDigest. These functions MAY be performed by the same implementation. References below to a content producer describe that function, not an additional trusted principal. The construction distinguishes (1) what the Issuer signed, (2) membership of disclosed records in that commitment, (3) permitted attribution and typed metadata, (4) integrity of supplied evidence, (5) agreement under a supported comparison, and (6) the relying party's application decision. A successful check at one stage MUST NOT be substituted for another stage's result.

Content commitment and explicit null For profile 0.7, engagement.contentDigest is REQUIRED. Its value MUST be either JSON null or sha256: followed by exactly 64 lowercase hexadecimal digits representing the 32-byte root defined in . An absent member, a different JSON type, an uppercase digest, or another digest scheme MUST NOT be interpreted as null or as a supported content commitment. A null value means that this statement binds no retained-content root. It does not assert that no evidence existed, that no fact was known, or that the physical engagement had no relevant content. A relying party MAY require a non-null commitment through an explicitly selected policy. A non-null value is not the JSON-DIGEST of the complete block document and is not an evidence-object digest. It is the separately domain-separated root specified below. It does not replace engagement.evidenceDigest, attestation.sealedEvidence.digest, site.envelope.digest, or chain.hash. Changing engagement.contentDigest changes the payload and therefore requires the existing payload-hash and signature procedures to be applied again. Null, a committed empty block, and a nonempty commitment whose content is not presented are different cases. An unbound presentation supplied with a null statement MUST NOT manufacture a root, vocabulary, fact count, or absence conclusion on behalf of that statement.

Content data model The construction identifier is the exact ASCII string pask-local-content-tree/1. The word local is part of the identifier and its hash input; it is not permission to substitute a local variant. This working specification proposes the implemented identifier without silently renaming its byte domain. The supported scope is the exact ASCII string PRESENTED_AT_SEAL. It records the Issuer's claim about the material committed when it sealed the statement. It neither authenticates an observation's time nor proves that a source actually supplied the material at that instant. A content header contains this construction identifier, this scope, and an exact vocabulary digest as defined in . The fact count is also bound in the header hash. A block contains zero through 256 fact records, each with its own 32-octet salt. A salt is separate from the fact JSON. Each fact record MUST be a UTF-8 JSON object with the following members:

Member	Requirement
name	REQUIRED nonempty string, at most 128 ASCII bytes, drawn only from lowercase letters, digits, dot, and hyphen.
value	REQUIRED; its admitted type and values are determined by the selected vocabulary row. Presence is distinct from null.
assertedBy	REQUIRED string identifying a recorded attribution category.
basis	REQUIRED string identifying the recorded basis category.
unit	OPTIONAL; if supplied it must satisfy the selected row as specified below.
evidence	OPTIONAL unless required by the row or the global party rule; its allowed forms are defined in .
remoteOrigin	Required or forbidden exactly as stated by the selected row.

Additional members of the outer fact object are retained and included in the commitment. They MUST NOT be stripped, rewritten, or silently endowed with verification semantics. A value object is closed only where its selected schema explicitly says so. In particular, the closed site.pose value does not retrospectively close the older fact objects. Fact names MUST be unique within a block. The producer MUST order facts by ascending ASCII byte order of their names and assign zero-based indices in that order. One name identifies one committed slot; this construction does not encode a time series by repeating a fact name. The complete fact bytes MUST equal their JCS serialization under . A verifier MUST reject malformed or noncanonical fact bytes, including duplicate JSON member names, without normalizing them for a passing proof. Canonicalization of prepared source data happens before commitment. Original evidence bytes are governed separately by and MUST NOT be canonicalized merely to make a digest agree. The construction uses the finite binary64/JCS number model. Exact integer-form inputs (without fraction or exponent) are restricted to [-9007199254740991, 9007199254740991]. Other admitted canonical finite binary64 forms do not establish exact large-integer provenance. A value outside the supported numeric domain MUST NOT be rounded into it. A canonical finite non-integer can be committed but may violate its vocabulary row. The present vocabularies impose integer rules on their numeric value fields; that restriction is not a claim that RFC 8785 prohibits all fractional numbers. Parsing cannot reconstruct precision already lost upstream, and a report MUST NOT infer a source's original numeric spelling from its canonical representation. For this bounded construction, a fact is at most 16384 octets, aggregate fact bytes are at most 1048576 octets, and fact JSON container nesting is at most 16, counting the outer container as one. No membership proof has more than eight sibling hashes. Out-of-bound material MUST NOT be reported as successfully checked. Limits are not evidence of physical completeness, constant memory use, constant time, or bounded acquisition outside the API. A conforming producer MUST use fresh, independently generated, unpredictable 32-octet salts and MUST NOT reuse a salt within a block. It SHOULD avoid reuse across blocks and protect undisclosed salts with the retained content. The implementation checks salt length and reuse among material it can see; it does not test entropy or establish secrecy or freshness. Deterministic test salts do not demonstrate production salt generation.

Byte-exact content hashing All hashes in this subsection are SHA-256 over octet strings. || denotes concatenation; U64(x) is an unsigned 64-bit, big-endian encoding; len(x) is an octet length; and B(t) is the ASCII encoding of a stated literal. Hexadecimal notation below denotes octets, not printable hex characters. Let D be ASCII PASK-LOCAL-CONTENT-TREE/1 followed by one zero octet. Let C = B("pask-local-content-tree/1"), S = B("PRESENTED_AT_SEAL"), V be the 32 raw digest octets decoded from the vocabulary digest, and N be the fact count.

H = SHA-256(D || 0x00 || U64(N) || U64(len(C)) || C || U64(len(S)) || S || V) L_i = SHA-256(D || 0x01 || H || U64(i) || salt_i || U64(len(F_i)) || F_i) Empty(H) = SHA-256(D || 0x02 || H) Node(A, B) = SHA-256(D || 0x03 || A || B) Root = SHA-256(D || 0x04 || H || Tree(L_0, ..., L_(N-1)))

F_i is the entire canonical fact record. Tree is defined recursively: the empty tree is Empty(H); a one-leaf tree is that leaf; otherwise split the leaf sequence at the largest power of two strictly less than its length and hash the left and right subtree roots with Node, in that order. The result for engagement.contentDigest is sha256: plus the lowercase hexadecimal encoding of Root. The construction binds the literal construction/scope, vocabulary digest, count, indices, salts, lengths, and complete disclosed fact bytes. It does not by itself bind a site or session; the PSER signature associates the root with the surrounding payload. This is a content-membership construction, not the SCITT log-leaf hash. Its domain prefixes and root wrapping MUST NOT be replaced with a Transparency Service's RFC9162 or CCF leaf procedure. The separate candidate-entry rules in remain unchanged.

Disclosure data and membership verification A disclosure presentation supplies a header, a count, and zero or more disclosures. Each disclosure supplies the original canonical record bytes, its 32-octet salt, index, and a sequence of 32-octet sibling hashes ordered from leaf to root. Left/right placement is derived from the index, count, and recursive split above; a caller-provided direction flag is not used. The data model is independent of its transport. An encoding MUST preserve the exact octets, integer values, and distinctions specified here. describes the reference implementation's local file container; that container is not a SCITT media type or a public registration protocol. Before asserting statement-bound content, a recipient MUST verify the statement using its selected key and the applicable profile's payload/header checks, and obtain the expected root from that checked payload. An unrelated caller root, an unprotected-header hint, or a presentation's own calculated root MUST NOT substitute for engagement.contentDigest. Key-to-identity trust remains an additional question. For each disclosed fact the recipient MUST check bounds, canonical bytes, fact shape and name, index range, salt and hash widths, exact proof consumption, and the reconstructed root. It MUST reject duplicate disclosed indices, duplicate disclosed names, reused disclosed salts, and names whose ordering disagrees with their indices. It MUST NOT silently discard an invalid disclosure to turn the remaining batch into an apparently complete presentation. A resource failure or unsupported construction is reported as such, not converted into missing evidence. A nonempty block with no disclosures supplies no verified header or count. The result is not presented, not proof of absence. For a claimed empty block, the recipient can recompute the header and empty root and compare it with the expected root. A correctly proved empty block is distinct from null. All committed slots are presented only when every in-range index is accounted for once, all proofs match, and the disclosed name/salt/order constraints hold. A selected subset proves only its disclosed slots. It cannot validate hidden records, prove that hidden names or salts are unique, or establish that the full block obeyed the producer rules. Membership and semantic checks are independent. A record with a forbidden attribution or invalid typed value can still be correctly committed. Such a record MUST retain its adverse finding rather than be made unprovable or silently omitted from the recipient's report. A producer or recipient MUST NOT describe it as a conforming typed fact merely because the hash or signature matches.

Versioned content vocabularies This document defines two closed, versioned rule sets in . Their exact identifiers and source-file digests are listed there. Vocabulary version numbers are independent of the PSER profile, implementation package version, and Internet-Draft revision. The vocabulary digest is SHA-256 of the exact published UTF-8 JSON artifact bytes, including their retained formatting and terminal newline. It is NOT the JSON-DIGEST of a newly serialized copy. The committed field uses sha256: plus that hash. Exact artifacts accompany the specification for reproduction; all rule definitions are also included in so implementing the rules does not depend on access to a private repository, a service, or a moving branch. A recipient MUST select a supported table by the exact digest from the verified content header. It MUST NOT select by display version alone, by a recognized fact name, by an unbound header, or by a default. An unknown digest is unsupported even when the content membership itself verifies. Compiled tables MAY be used when they reproduce the specified rules for that exact digest. A recipient need not fetch a vocabulary at run time. The selected table governs all fact-dependent operations: name support, party/basis classification, forbidden pairs, typed values, implicit units, evidence requirements, remote-origin metadata, and required-name support. Each block in a mixed presentation MUST select independently. A later table MUST NOT reinterpret an earlier commitment. The /1 table has 46 facts. The /2 table retains their rule definitions and adds site.pose. The existing /1 digest identifies the spelling-migrated table reproduced by the implementation; another historical file sharing its display version does not acquire this digest's meaning.

Attribution categories and permitted relation assertedBy records a provenance category claimed in the fact; it is not a party-specific signature, credential, or authentication mechanism. The seven category labels and their global bases are:

Category	Global permitted bases
appliance-measured	measured, estimated
robot-attributed	measured, estimated, declared
operator-entered	estimated, declared
platform-recorded	measured, declared
site-policy	declared
manufacturer-declared	declared
undetermined	estimated, declared

The three basis values are exactly measured, estimated, and declared. Their presence is the Issuer's recorded claim; neither a plausible label nor a permitted combination establishes how the value was obtained. In particular, robot-attributed MUST NOT be reported as robot-signed. A manufacturer-declared limit and a robot-attributed limit remain different recorded provenance even when both are permitted and both are declared. For a known fact, known party and known basis, the combination is permitted if and only if all of the following hold: the row admits the party, the row admits the basis, the party's global list admits that basis, and the pair is not in the row's pairsForbidden. No row overrides a global exclusion. Unknown fact, party and basis are distinct unsupported classifications, not a known forbidden attribution. This relation classifies provenance choices only. Typed-value failures, missing required evidence, malformed references, unit failures, resource failures, and origin-metadata failures remain separate findings and MUST NOT be erased by a permitted relation.

Value dialect and additional metadata The row schemas use the following finite dialect; they are not a claim to implement a general JSON Schema vocabulary.

• type is one of boolean, integer, string, object, or map-of-integer.
• nullable: true admits JSON null; otherwise null is invalid. It does not make the containing member optional.
• Integer values have no fractional/exponent interpretation at the typed layer, use signed 64-bit bounds, and obey inclusive minimum and maximum where present. The narrower canonical-content number domain above still applies to committed records.
• String enum compares exact decoded text, without case folding or Unicode normalization. maxLength counts Unicode scalar values.
• The only patterns are a nonempty ASCII identifier consisting of letters, digits, colon, underscore, dot and hyphen; the lowercase sha256: digest spelling; and the fixed 24-character millisecond-UTC textual shape. The last is a lexical check, not a calendar or clock check.
• An object has the named properties and required members in its row. Unknown properties remain permitted unless the /2 row explicitly has additionalProperties: false. Present known properties are checked recursively. map-of-integer requires all member values to be integers.
• Object/map width is at most 256 and recursive schema-value evaluation has root-zero depth at most eight in this bounded dialect.

An implementation interpreting a /2 rule definition MUST NOT treat an unrecognized schema keyword, an invalid keyword type or combination, or an unsupported pattern as satisfied. It MUST report the rule definition as unsupported or invalid, as applicable, rather than silently ignore it. Rejecting such definitions during code generation is one way to enforce this requirement. This does not retrospectively change /1's generation recipe or its open object behavior. The exact schemas in are the supported definitions; arbitrary schema extensions require a separately specified vocabulary. The row's unit supplies the semantic unit even when an optional fact-level unit member is omitted. If supplied, that member MUST be the matching unit string, or null only where the row is unitless. Malformed types or a different unit MUST remain invalid. Omission does not promote a unit-bearing fact into a unitless comparison. remoteOrigin is control-origin metadata, with values on-site, off-site, and undetermined. It is REQUIRED exactly for rows marked required and is forbidden, including explicit null, elsewhere. The tag does not authenticate acquisition location or source identity.

Position observations (site.pose)

Meaning and optionality The site.pose fact represents one position observation of one site-declared reference point from one source sample. Despite the fact name, this representation does not include orientation, heading, speed, trajectory, or an orbital ephemeris. Presence of site.pose is not a universal requirement for a receipt or for every mobile site. A separately supplied disclosure policy can require its presentation. The policy's origin and authority are separate from whether its stated disclosure requirements have been met. The fact is retained and disclosed through the content-block mechanism. A witness signature that binds the content commitment binds the recorded observation and its attribution to that statement. It does not establish that the site occupied that position, that the named party produced the observation, or that the measurement is accurate. One site.pose record is permitted per block under the existing unique-fact-name rule. This vocabulary entry does not define a time series.

Value representation The value MUST be a closed object containing all eight members below. Additional members MUST produce an invalid-typed-value finding; they MUST NOT be removed or rewritten to make the value pass. Only the three explicitly nullable members MAY be JSON null.

Member	Type and admitted values	Meaning
crs	String, exactly EPSG:4979 ()	Identifies the coordinate reference used by this scaled encoding.
referencePoint	String, 1–64 characters matching ^[A-Za-z0-9:_.-]+$	The site-declared point actually positioned by the source.
latE7	Integer, -900000000 through 900000000 inclusive	Geodetic latitude in units of 10^-7 degree.
lonE7	Integer, -1800000000 through 1799999999 inclusive	Geodetic longitude in units of 10^-7 degree; the representation uses -180 degrees rather than +180 degrees.
heightMm	Integer, -20000000 through 100000000 inclusive, or null	Ellipsoidal height in millimetres, or no ellipsoidal height asserted.
horizontalAccuracyMm	Integer, 0 through 100000000 inclusive, or null	The coherent source sample's reported accuracy figure, or no such figure asserted.
observedAt	Millisecond-UTC string of the specified 24-character shape, or null	Source-reported sample instant, or no source instant asserted in this member.
latLonDerivation	String, exact-integer-rescaling or rounded-half-even-to-1e-7-deg	The stated preparation of the horizontal pair only.

The coordinate meaning is latitude = latE7 / 10^7 degrees, longitude = lonE7 / 10^7 degrees, and, for non-null height, ellipsoidal height = heightMm / 1000 metres. This is a scaled encoding referencing the coordinate system, not a change to its units. The row supplies neither a datum realization nor a coordinate transformation. Null height means that height is not asserted. It does not mean zero, a complete three-dimensional observation, or a conversion to another coordinate system. Sea-level, chart-datum, or geoid-relative height is not interchangeable with ellipsoidal height. The height and accuracy ranges are selected support limits for this row, not universal physical or geodetic validity rules. Integer resolution does not establish physical accuracy. Literal zero is a value wherever admitted. The string unknown is not a replacement for null. A null member inside a disclosed site.pose value is distinct from a null content commitment (no content root bound by this statement), from a withheld fact, from a name the block's vocabulary does not support, and from a fact absent from a complete block. The definition of referencePoint belongs in the retained envelope document. This row does not require or implement resolution of that definition by the current recipient. The accuracy figure's confidence level and metric are unspecified. The figure does not establish a verified radius, a 95-percent bound, cross-source comparability, or an automatic policy result. Preserve its source meaning in evidence; do not invent confidence metadata.

Preparation obligations and what the checker can establish All populated components MUST describe one coherent source sample and reference point. An unrelated accuracy sample does not supply the accuracy of this position. A different reference point is not substituted without an explicitly described transformation; no such transformation is introduced here. Before commitment, the producer MUST process horizontal source values in exact decimal arithmetic. Source latitude outside [-90, 90] degrees or longitude outside [-180, 180] degrees is rejected rather than clamped. Multiply each by 10^7. If both products are integers, use exact-integer-rescaling; otherwise round the horizontal pair by round-half-even, including negative ties, and use rounded-half-even-to-1e-7-deg. Write a rounded zero as canonical 0; encode +180 degrees as -180 degrees. Verification does not repeat these transformations on presented fact bytes. A coherent populated height or accuracy value MUST rescale exactly to integer millimetres and lie inside the row's supported range. For example, 12.345 m and 12.3450 m both yield 12345 mm exactly. A coherent known 12.3456 m value, or a coherent value outside the supported range, is not rounded, clamped, or silently reclassified as missing, and the producer MUST NOT emit that candidate site.pose. It MUST retain the raw source and record the not-representable condition in its acquisition or mapping record. Actual missing information, a different-datum height, or an incoherent accuracy sample is different and does not supply the member being asserted. The producer MUST NOT default observedAt from the receipt's asserted sealing time, local acquisition time, or registration time. A source instant exactly expressible in UTC milliseconds can be represented without loss, including whole seconds as .000 and an equivalent .123000 fraction as .123. Nonzero sub-millisecond precision, unknown offset, or an unapplied non-UTC time scale is not silently rounded or truncated: use null and retain the original time and the reason outside that member. This does not assert that the source lacked a timestamp. The admitted lexical shape is YYYY-MM-DDTHH:MM:SS.mmmZ. The current implementation checks this shape, not calendar validity or real-world clock correspondence. A string matching the pattern can still denote an impossible date or time. These preparation and coherence rules are producer obligations in the proposed contract. The accepted implementation validates the resulting representation; it does not contain a navigation-source converter, a calendar validator, or a check that the claimed transformation actually occurred. Do not report those absent checks as executed functionality.

Attribution, evidence, and control-origin metadata The following four party/basis pairs are admitted:

assertedBy	Permitted basis
appliance-measured	measured, estimated
platform-recorded	measured
operator-entered	declared

The other seventeen combinations of the seven existing parties and three bases are forbidden for this fact. In particular, platform-estimated is a forbidden known combination, not an unknown-vocabulary result. These restrictions do not widen any global rule. The row's optional evidence setting does not waive the global evidence-reference requirement for appliance-measured. A supplied reference, available object, matching raw-byte digest, compared value, and authenticated party are separate findings. The existing fact-level remoteOrigin member MUST NOT appear on site.pose, including with a null value. This prohibition concerns that control-origin tag. It does not establish on-site acquisition and does not forbid retaining off-site navigation evidence. Acquisition origin remains in the retained evidence or mapping record; this row adds no new signed acquisition-origin member.

Vocabulary selection and recipient reporting The exact vocabulary digest in the proof-bound content header selects the supported rule table. A version label, fact name, unbound header, or caller hint is not a substitute for that digest. Unknown digests do not fall back to the older table. The selected table governs all fact-dependent checks: party/basis rules, value schema, evidence requirements, origin metadata, implicit units, and the support of required fact names. Blocks in a mixed presentation are evaluated using their own committed vocabularies. Existing v1 semantics and commitments remain unchanged. The reference implementation accepts a policy name recognized by either supported vocabulary, then determines support within each block. A name unknown to both remains an unsupported policy configuration. The following availability findings are distinct:

Situation	Reference-implementation result
Required site.pose under a proved v1 block	UNSUPPORTED_BY_VOCABULARY; not policy satisfaction, not a claim of absence, not a forbidden attribution; the disclosure-policy result remains unestablished.
Required pose not disclosed by a v2 subset	NOT_DISCLOSED_UNPROVEN.
Pose absent from all slots of a complete v2 block	ABSENT_FROM_COMMITTED_BLOCK; no claim about physical-world absence.
Unknown vocabulary digest after valid membership	Vocabulary support is Unsupported; required-name availability is NOT_EVALUATED. An unbound header leaves table-dependent checks unevaluated. No table is chosen by label, fact name, or default.
Null commitment	Presentation is NULL_COMMITMENT; a requested fact is NO_BLOCK_COMMITTED. This statement binds no root; it does not prove that no uncommitted block existed.
Non-null commitment, block not presented	Presentation is NOT_PRESENTED; a requested fact is COMMITTED_BLOCK_NOT_PRESENTED. The unbound header does not select a vocabulary.

A verified membership proof does not make an invalid value or forbidden attribution valid. A correctly matched evidence object does not become a verified position: the supported root-scalar comparator cannot compare a compound position object. Comparison enabled can therefore produce NOT_COMPARABLE; comparison disabled remains NOT_RUN. A satisfied disclosure policy does not erase contradictory evidence, metadata failures, forbidden attributions, or unavailable trust inputs. It is not application acceptance.

Evidence references and value comparison A fact's evidence member, when non-null and present, MUST be an object containing exactly one of:

{"digest":"sha256:<64 lowercase hexadecimal digits>"}

or

{"pointer":"<nonempty locator of at most 2048 UTF-8 octets>"}

The angle-bracketed strings above describe placeholders, not valid literal digest instances. The two members MUST NOT be combined in the same reference under this vocabulary. Absent or null evidence means no reference is given. A reference is REQUIRED if the row requires evidence OR the party is appliance-measured or robot-attributed. A row marked optional does not waive that global rule. A pointer can satisfy the reference-shape requirement while still supplying no immutable byte-integrity binding. To check evidence integrity, the expected digest MUST come from the proved fact, not from the supplied object's own label or from hashing that same object and treating the answer as an independent expectation. The recipient hashes original supplied bytes with SHA-256. It MUST NOT normalize JSON, transcode text, trim whitespace, or rewrite a numeric token before that hash. Acquisition and integrity are separate: not requested, unavailable, and present zero-length bytes are different states. A pointer alone, a returned HTTP body, a matching filename, or a successful retrieval MUST NOT be reported as a digest match. Any implementation that fetches data must define its retrieval authorization, limits, and trust assumptions separately; the local implementation here does not fetch. The local digest-keyed evidence inventory rejects duplicate digest keys rather than selecting the first or last copy, including identical duplicate bytes or mixed availability declarations. Malformed inventory data MUST NOT be converted into an apparently weaker but successful attribution-only finding. A recipient performing value comparison MUST identify the comparison rule, its required inputs, and the result independently of byte integrity. It MUST report a detected contradiction, and MUST NOT treat unsupported comparison as agreement. A matching digest does not establish relevance, truthful content, authentic origin, or consistency with the sealed fact value.

Bounded unitless scalar comparison The local reference comparator is json-root-unitless-scalar-exact/1. Invoking it is an explicit policy choice, not an assertion that every vocabulary row has this evidence encoding. This comparator runs only after a raw digest match and after applicable fact metadata checks. It supports a complete JSON root Boolean, text string, or exact signed/unsigned 64-bit integer, with no unit. It does not select a property from an object, transform units, compare a compound pose, infer numeric types from strings, or normalize Unicode. For this comparator, matching supported types and equal values give MATCH; different values of the same supported type give CONTRADICTION. Unsupported types, units, or type mismatch give NOT_COMPARABLE. A disabled comparison or failed prerequisite gives NOT_RUN. Original integer-token -0 is compared as zero after syntax validation and after hashing its original bytes; -0.0, fractional and exponent tokens do not become supported integers. JSON permits only its defined surrounding whitespace; invalid syntax remains not comparable. These comparator-specific constraints do not redefine the content canonicalization rules. A raw evidence object can have a different spelling from the recorded value; equal semantic values do not excuse a wrong raw digest.

Recipient results and disclosure policy A recipient MUST retain separate outcomes for statement inspection, content membership, vocabulary support, attribution/basis classification, typed metadata, evidence integrity, value comparison, requested-fact availability, and selected disclosure policy. There is no mandated universal valid or accepted Boolean. A relying party MAY reach an application decision, but MUST identify its policy and additional trust assumptions rather than present that decision as the output of a narrower cryptographic check. The reference implementation retains its existing inspection dimension: Passed, Failed, Unsupported, Unestablished, and NotEvaluated. Its content-fact provenance dimension is separate:

Provenance finding	Meaning
EVIDENCE_LINKED	A permitted, well-formed committed fact has matching original evidence bytes. This may coexist with a value contradiction.
ATTRIBUTION_ONLY	The permitted committed attribution lacks checked evidence bytes; the separately recorded reason remains visible.
ATTRIBUTION_FORBIDDEN	The known fact/party/basis combination violates its selected rule relation. Other adverse findings remain visible too.
FACT_UNAVAILABLE	The requested-fact result lacks a disclosed fact under its particular availability conditions; the precise availability reason controls its meaning.

An absent provenance classification is also meaningful. Unknown rules, malformed metadata, bad integrity or inventory, and unmet prerequisites MUST NOT be forced into ATTRIBUTION_ONLY or a physical absence claim merely to populate this dimension. The recipient obtains its expected root from the checked statement and verifies the presentation before applying a vocabulary. When any disclosure invalidates that presentation, it MUST NOT return a partially successful set of statement-bound fact reports. Independently known statement or framing results MAY remain visible. A disclosure policy can require content, all committed slots, or named facts. A requested name recognized by some implemented vocabulary but absent from the selected block's vocabulary is unsupported for that block; it is not a globally malformed policy and MUST NOT be silently satisfied. Names unknown to every implemented table can be rejected as unsupported configuration. With mixed-vocabulary input, support is decided per block. The following distinctions MUST remain observable:

Situation	Meaning
Null commitment	No root bound by this checked statement; no assertion about uncommitted material.
Non-null commitment, no usable disclosure	Committed content not presented; do not infer the hidden count or facts.
Known-vocabulary subset omits a requested fact	Nondisclosure; absence is unproven.
All committed slots under a supported vocabulary omit a requested fact	Absence from this particular committed block only.
Required name not defined by the proved vocabulary	Unsupported by that vocabulary, not forbidden attribution or proven absence.
Unknown vocabulary digest	Vocabulary unsupported; table-dependent required-name evaluation not evaluated.
Malformed disclosure or invalid proof	Invalid/unsupported presentation as applicable, not missing content.
Matching empty-block root	A proved empty committed block, not null.

Meeting disclosure requirements MUST NOT erase forbidden attribution, invalid metadata, integrity failure, value contradiction, or missing trust inputs. It does not establish semantic completeness or application acceptance.

Custody, concealment, and history The Site Owner is responsible for arranging retention and authorized retrieval of the committed records, salts, disclosure material, vocabulary artifacts, and referenced evidence needed for later examination. Storage MAY be delegated; delegation does not itself establish accessibility or change the declared custody responsibility. The applicable site policy SHOULD state retention, access, and deletion arrangements. This profile defines no universal retention period and grants no recipient a right to disclose private material. Loss, refusal, or inability to retrieve committed content MUST remain an availability finding. It MUST NOT be translated into null, rewritten into a different block under the original root, or described as proof that no fact existed. A malicious or compromised Issuer can sign null or omit a fact before commitment. Requiring content through policy helps detect policy nonfulfilment; it does not prove complete observation. For two or more statements presented as one chain, the existing Chain-Verifier rules apply: check adjacent sequence numbers and predecessor hashes in the presented order and retain affiliation changes at the affected records. This section introduces no replacement chain algorithm. Reordering, deduplicating, or filling missing entries MUST NOT be hidden as verification. A valid presented prefix does not prove latest or complete history. The content root adds neither a trusted closing head nor evidence that a later statement or event was not withheld. Reusing a block in two correctly signed contexts is not prevented by membership. A context comparison checks the provided signed labels against caller expectations; it does not authenticate their origin or prove same-site continuity.

Local replay implementation profile (informative) The reference workflow uses pask-local-recipient-replay/1, a local JSON container with entries holding original statement/fact/evidence bytes encoded as lowercase hexadecimal. Its supplied verification key is external to the document. Policy and expected context remain explicitly supplied, unauthenticated inputs. This container and its report schema are implementation interfaces, not new COSE parameters, SCITT media types, or mandated network protocols. SINGLE mode requires one entry and leaves chain contiguity unevaluated. CHAIN mode verifies its entries under the supplied Ed25519 key, then checks the presented sequence through the existing chain helper, including the selected genesis-at-zero requirement. A valid singleton genesis is accepted in that explicit mode. The workflow guards sequence overflow and neither sorts nor fills the presentation. It retains per-statement content findings separately from its chain result. It does not check a common site identity merely because the links pass. The executable uses exit zero to mean that a report was produced, not that every finding passed. Invocation, input/framing/resource, key, and ordinary I/O failures use exit two. A successful report can contain failed membership, contradiction, or unsupported rules. The tested output-failure behavior does not cover arbitrary panics, signals, blocking devices, or allocation failure. The reference limits include 8 MiB of raw replay JSON, depth 32, 16 entries, and 3 MiB of decoded binary material. Per-entry local evidence is limited to 256 objects, 64 KiB per object and 1 MiB in aggregate. That inventory can induce up to 16 MiB of repeated evidence hashing per entry and up to 256 MiB per 16-entry batch. Decoded size is not total allocation or work. These are implementation operating bounds, not a production performance guarantee. The local replay does not validate attached SCITT Receipts, appraise hardware, authenticate keys or policies, locate a signing key, or make an application decision. The separately implemented Receipt-verification coordinator is not silently invoked by this workflow.

Use with robot safety records (informative) Industrial robot safety standards distinguish the robot from its integrated application and call for application-specific risk assessment, verification, validation and documentation. ISO 10218-1:2025 , 5.3.5, and ISO 10218-2:2025 , 5.5.5, address identifying safety-function settings. ISO 10218-2:2025, Annex N.2.4.4, provides informative guidance for retaining measurement conditions and the parameter identifier used in testing. ISO/TS 15066:2016 provides related guidance for collaborative industrial applications; its dated references and terminology remain edition-specific. The retained-content mechanism can support exchange of recorded claims about a safety-parameter digest, operating context and associated test or observation evidence. In unit.safety-parameter-digest, "active" is part of the attributed claim, not a recipient-verified device state. The manufacturer's native identifier, a SHA-256 digest over exact retained export bytes, the export's stated coverage, the claim that those settings were active, and any validation record are separate items. An application identifies the represented data, algorithm, coverage and provenance rather than assuming interchangeability. The digest stored as this fact's value is distinct from a digest in its evidence reference. The current evidence-integrity check compares the original supplied bytes with the evidence-reference digest in the proved fact; it does not compare that digest with the recorded value. Evidence can be an export or a report about an export, so digest equality is not an implicit rule for every evidence relationship. For example, a fact can record a digest naming configuration X while its evidence reference correctly binds a different configuration export Y. With that compound export as evidence, the current recipient can report MATCHED and EVIDENCE_LINKED together with NOT_COMPARABLE. Those findings do not establish that Y describes the recorded digest, that the settings were complete or active, or that changes were authorized or safe. They do not authenticate the attributed party. An application that needs a particular configuration-to-evidence relationship checks that relationship separately; it is not supplied by the byte-integrity result. The vocabulary is not an ISO safety classification or conformity scheme. Its attribution categories, contact labels and mode labels have the meanings defined in this profile. Any mapping to a safety standard states the relevant edition and application assumptions. A single operating-regime label need not describe all simultaneously active protective measures. A successful signature, disclosure proof, evidence digest or supported value comparison does not establish that settings were active or safe, that measurements were calibrated, or that a robot application conforms to a safety standard. In particular, the current unitless scalar comparator does not assess unit-bearing force values, measurement traces, pressure, stopping performance or biomechanical criteria. Recorded configuration, measured outcome and selected safety limit remain distinct. Round-trip network latency is not a safety-function response or stopping time. A site.pose observation is not a tool-centre-point pose or a human-to-robot separation measurement. Retained content can assist later examination of evidence; it does not replace the safety functions, application risk assessment or validation activities. The applicability of industrial-robot standards is determined for the deployment. Their principles can inform a non-industrial example without making that example a conformity test.

SCITT registration and Receipt attachment A Physical-Site Engagement Receipt Signed Statement is registered with a SCITT Transparency Service per Section 6.3. The TS applies its Registration Policy against the protected header (in particular iss, sub, and content_type) before registering. Upon successful registration, the TS returns a Receipt as defined in . The Receipt is attached to the Signed Statement's unprotected header as an element of the receipts array (CBOR label 394), producing a SCITT Transparent Statement per Section 7. The same Signed Statement MAY be registered in multiple Transparency Services and MAY carry multiple attached Receipts, one per Transparency Service, per Section 6.3. Registration is mandatory in this profile. An Issuer MUST register every Physical-Site Engagement Receipt it issues with at least one Transparency Service. A relying party MUST NOT accept a Physical-Site Engagement Receipt as conforming to this profile unless at least one attached Receipt from a Transparency Service that relying party trusts verifies per . Verifying an attached Receipt does not demonstrate that the Issuer registered every receipt it issued; a relying party that requires that assurance MUST obtain it from the Transparency Service's own audit and consistency mechanisms, not from an individual attached Receipt. Requiring registration does not require a relying party to be online when it verifies. An attached Receipt is a Verifiable Data Structure Proof per , checkable from the presented bytes together with the Transparency Service's verification key, both of which MAY be held locally. The offline-verifiable property stated in is preserved: what registration adds is a reference obtained before verification, not a network dependency during it. This revision defines no conforming mode of operation in which no Transparency Service is reachable at issuance time. This revision specifies the candidate-entry byte encoding for inclusion-proof verification. For this profile, the presented object MUST be a COSE_Sign1 with four array elements: protected-header byte string, unprotected-header map, attached-payload byte string, and signature byte string. The payload MUST be present as a byte string; a null payload indicating detached content is not permitted. The input may be an untagged COSE_Sign1 or a COSE_Sign1 wrapped in tag 18. Other tag wrappers are not accepted by this derivation. The verifier derives the candidate entry as the untagged four-element array [P, {}, M, S], where P, M, and S are the original protected-header, payload, and signature byte-string contents. The candidate entry MUST use the core deterministic encoding requirements of , Section 4.2.1, for the outer array, empty map, and byte-string framing. The contents of P, M, and S are preserved unchanged; this requirement does not instruct the verifier to parse and reserialize those contents. The outer array is definite-length, the empty map is encoded as 0xa0, and the three byte strings use the shortest definite-length encodings of their lengths. This derivation MUST preserve P, M, and S by content. It MUST NOT parse and reserialize their contents to construct the candidate entry. The original outer CBOR length encodings and any permitted outer tag are not preserved. These derivation rules do not waive separate COSE header, signature, payload, or profile validation. Parsing a header for validation is distinct from rewriting its bytes for hashing. This follows RFC 9943 Section 6.3, which requires the unprotected header of a Signed Statement to be set to an empty map before inclusion in the Statement Sequence. Transmitted envelopes use COSE tag 18 (), and attached receipts are encoded as byte strings containing tagged Receipt objects per Section 4.3. The untagged candidate entry is a profile-internal representation used for hashing; it is not the transmitted wire form. A Transparency Service that hashes a different representation (for example, the tag-18 form) cannot satisfy this profile unless it commits to the candidate-entry representation specified here. This is the profile imposing a requirement on a party it does not control: an existing TS that hashes the tagged form cannot satisfy the profile without adaptation. A Transparency Service used to satisfy this profile's inclusion requirement MUST issue a Receipt whose inclusion proof commits to the candidate-entry representation specified in this section. The applicable registration path and the verifier MUST use the same derivation and serialization rule. Emptying the unprotected map alone does not establish agreement on the remaining serialized bytes. The service's internal storage format is not prescribed, but the candidate-entry bytes against which its proof verifies are prescribed. Registration and verification alignment MUST be demonstrated by byte-exact fixtures, including derivation from the presented Transparent Statement without access to retained producer-side bytes. Attaching, removing, or modifying receipts in the unprotected header does not change the candidate entry: candidate_entry(S) == candidate_entry(attach_receipt(S, R)) for any receipt R. This is the property that makes inclusion proofs portable: the proof covers the registered statement, not whatever receipts happen to be attached to it at verification time. The profile's inclusion requirement is satisfied only when at least one attached Receipt has both a valid Transparency Service signature under an accepted service key and a valid inclusion proof for the derived candidate entry. A Receipt that is invalid, unsupported, or associated with an untrusted service does not count toward that requirement. A structurally parseable additional Receipt whose cryptographic verification fails MUST NOT, solely by its presence, defeat another attached Receipt that satisfies the inclusion requirement. A separately identified local policy may impose stricter acceptance conditions. This rule does not relax validation of the enclosing COSE structure or malformed attachment containers. Receipt signature verification, inclusion-proof verification, and issuer-signature verification are separate checks. Satisfying the inclusion requirement does not establish full PSER conformance; every other applicable profile requirement still applies. An absent proof and an invalid, unsupported, or untrusted proof are not reported as successful inclusion. For RFC9162_SHA256, the leaf hash is SHA256(0x00 || candidate_entry), using the leaf construction in , Section 2.1.1. Receipt representation and verification follow the applicable RFC9162_SHA256 procedures in . This formula is not applied by default to another or unknown VDS. Verification of another supported VDS uses that VDS's specified procedure. This paragraph does not add a PSER-wide prohibition on other SCITT VDS types. The current reference implementation's support is limited to RFC9162_SHA256; unsupported VDS values are reported as unsupported and cannot count as successfully verified evidence. This specification does not prescribe how a Transparency Service stores the entry internally. Interoperability requires only that the inclusion proof verifies when the verifier uses the candidate entry derived from the presented Transparent Statement.

Registration with an affiliated Transparency Service An Issuer MAY register with a Transparency Service it operates itself, or that is operated by a principal affiliated with it. Where it does so, the Issuer MUST disclose that relationship as an Issuer-published fact resolved under , and a relying party MUST NOT treat such a registration as evidence obtained from outside the Issuer for the purposes of . Where the disclosure does not resolve, the standing of the registration is undetermined; a Verifier MUST NOT resolve it to the unaffiliated case, which is the Issuer-favourable one. That obligation is separate from the issuerAffiliation member of Section 4.1, which states the relationship between the Issuer and the Site Owner rather than between the Issuer and the Transparency Service; neither can be inferred from the other. Registration with a Transparency Service operated by an unaffiliated principal is the only case in which an attached Receipt supplies a reference external to the party whose completeness is in question. This profile does not prohibit the affiliated case, because a self-operated Transparency Service still binds the Issuer to a consistent published history and still admits third-party auditing; it requires that the weaker standing of that case be visible rather than implied.

IANA considerations This document requests the following IANA actions.

Media type registration Register application/pser+json per , with the required profile parameter identifying the applicable PSER profile version. This document specifies the value wilder.pser/0.7. The values wilder.pser/0.5 and wilder.pser/0.6 identify the earlier profiles and retain their requirements. Unsupported values do not become supported by looking like a version number. A version-shaped value does not by itself identify a supported profile. Unknown or unsupported values are reported as unsupported and are not interpreted using the rules of another version.

COSE Header Parameters This document does not register new COSE header parameter labels. It uses only labels defined in , , and .

Content vocabulary identifiers No new IANA registration is requested for the content construction, document-defined fact names, or vocabulary labels in this revision. The exact digest dispatch and closed rule definitions are specified in and . They create no uncontrolled registry or wildcard interpretation rule.

New IANA registries This document requests the establishment of the following registries under a new "SCITT Physical-Site Engagement Receipt Profile" registry group, with policy "Specification Required":

1. Site Class -- values of site.class. Initial values: residential, industrial, healthcare, infra, other.
2. Engagement Type -- values of engagement.type. Initial values: patrol, service, inspection, delivery, installation, maintenance, presence.
3. TEE Class -- values of attestation.teeClass. Initial values: intel.tdx, amd.sev-snp, arm.cca, nvidia.h100-cc, nvidia.jetson-thor-cc, aws.nitro-enclave. These descriptions identify the environments named by the existing values. They do not, by themselves, define a complete evidence-format binding or establish implementation support.
   • intel.tdx: Intel Trust Domain Extensions, a VM-level isolation primitive.
   • amd.sev-snp: AMD Secure Encrypted Virtualization with Secure Nested Paging, a VM-level isolation primitive.
   • arm.cca: Arm Confidential Compute Architecture, a hardware isolation framework.
   • nvidia.h100-cc: NVIDIA H100 in Confidential Computing mode, a device-level isolation primitive.
   • nvidia.jetson-thor-cc: existing product-named identifier associated with NVIDIA Jetson Thor. Retained provisionally and unchanged; this profile does not establish the exact confidential-compute capability, evidence format, or appraisal binding represented by this value.
   • aws.nitro-enclave: AWS Nitro Enclave, a cloud-provider isolation primitive.
   A confidential-compute environment absent from the admissible set is not conforming and is not silently folded into an existing value. The registry is the route by which one becomes conforming, and its governance is stated above. This revision clarifies existing identifiers with descriptive context. It does not add, remove, or rename any value, and it does not resolve the evidence-format question for any value.
4. Sealed Evidence Encoding -- values of attestation.sealedEvidence.encoding. Initial values: opaque/1.
5. Operations-Layer System -- values of adapter.system. New values follow a vendor.product lowercase snake_case naming convention.

Security considerations

What this profile does NOT attest Per and the NORMATIVE non-goals stated there, a Physical-Site Engagement Receipt does NOT attest that:

• The engagement was safe, correct, effective, or compliant with any specific regulation.
• The site conditions were as recorded.
• No unrecorded engagement occurred outside the instrumented boundary.
• The operations layer targeted by the Adapter Write-In will use, act on, or preserve the receipt correctly.

Relying parties MUST NOT infer these claims from a receipt.

Verification scope and trust inputs Successful signature verification establishes a relationship between the presented signed bytes and the verification key used for that check. It does not, by itself, establish who controls that key, that a named device holds it, that the device was at the Site, or that the recorded content is true. Trust in the key and its association with a named role comes from the relying party's independently established policy and authenticated evidence, not solely from a key, identifier, certificate, or policy supplied alongside the statement. A credential conveyed with a statement can participate in such validation; its presence does not make it its own trust anchor. The direct-witness and delegated-witness requirements in state the relationships needed for the profile's hardware-related claims. A class label, a matching identifier, or a successful ordinary signature check does not establish those relationships. The reference implementation's current appraisal and trust-establishment limits are described in .

Equivocation and tail-truncation The chain field defined in makes in-band tampering detectable: modification, substitution, reordering, or omission of receipts interior to a presented chain breaks a chain.prevHash link, chain.seq contiguity, or a signature. This property holds against parties that do not hold the Issuer's signing key. An Issuer that holds the key can sign an alternative, internally consistent chain omitting receipts at any position. The chain field does NOT detect tail truncation -- the withholding of the most recent receipts -- in any presentation. A truncated chain is internally consistent at every link, and no property of the presented bytes reveals the withholding, because no receipt commits to a successor that did not exist when it was signed. This is not a limitation of the hash or signature algorithms: the presented bytes are identical whether or not a suffix exists. The chain field likewise does NOT detect equivocation, in which an Issuer signs two divergent chains for the same Subject. Detecting either condition REQUIRES evidence obtained from outside the presentation. Registration of a Signed Statement in a SCITT Transparency Service supplies such evidence to relying parties and auditors that check against that Service. Registration does not by itself establish completeness: a conforming Transparency Service does not compel an Issuer to register every Signed Statement it issues (, Section 9.3), and a Receipt proves the inclusion of one Signed Statement rather than the absence of others (). A Transparency Service therefore does not detect these conditions itself; it supplies the reference against which other parties can. A relying party that retains the highest chain.seq receipt it has verified for a chain holds such a reference. A later presentation whose head precedes that receipt, or which presents a different chain.hash at that chain.seq, is evidence of truncation or equivocation relative to it. Relying parties SHOULD retain these anchors. Detection reaches only as far as the anchor's own age: a presentation ending after the retained anchor is not thereby shown to be complete, and a presentation ending before it is not by itself proof of misbehaviour, since it may be an earlier honest observation. This profile does not define what a relying party does upon detecting such a mismatch, how long anchors are retained, or what evidentiary weight a mismatch carries. Those are matters for the relying party's own policy. A relying party that reproduces this section in a contract, underwriting rule, or adjudication SHOULD state its own remedy; this document supplies a detection property, not a remedy.

Corrective statements This revision does not define an interoperable corrective-statement wire format or impose correction-related requirements on 0.5, 0.6, or 0.7 conformance. An informative corrective-statement design is recorded in for discussion. A correction does not reset the original's attestation.validity interval. Conflicting corrections are surfaced, not resolved. No automatic winning history is computed.

Adapter Write-In is write-only in this revision The Adapter Write-In records that the receipt was posted into an operations layer. It does NOT permit the operations layer to write back into the receipt or the TEE. The adapter.mode field is fixed to WRITE_ONLY in this revision; a future revision MAY define a WRITE_READ mode with additional security machinery. Implementations that reverse this direction in a way that permits the operations layer to modify Issuer or TEE state are NOT conforming to this profile.

TEE compromise A compromised TEE can produce receipts that are cryptographically valid under this profile but describe engagements that did not occur or did not occur as described. Detection of TEE compromise is out of scope of this profile and depends on the platform-native attestation supply chain identified by attestation.teeClass. Relying parties SHOULD consult Section 9 for guidance on Issuer participation and key management, and the TEE vendor's own security guidance for the specific teeClass.

Witness key lifecycle A witness key signs inside a TEE that is physically hosted at a Site the Issuer may not control. The parties that can act on a suspected compromise of such a key are therefore not the same as those that can act on a compromise of a key the signer holds itself, and this document states which party may make which assertion. This revision addresses assertions about a specific witness key. Compromise of a TEE class or platform is addressed in and is not a key lifecycle event under this section.

The two assertion classes This document defines two distinct assertions about a witness key. They are named rather than numbered so that a later revision may define a third without redefining either.

• Cessation. An assertion that the identified witness key MUST NOT be relied upon to produce further receipts. Cessation is forward-looking only.
• Retroactive impeachment. An assertion that receipts already produced by the identified witness key SHOULD NOT be relied upon, in whole or over a stated interval. Retroactive impeachment reaches backward, and it is the stronger of the two.

Authority over each is asymmetric, and the asymmetry follows the capabilities the trust model already grants in :

• Cessation MAY be asserted by the Site Owner or by the Issuer, independently of one another. Neither party requires the other's concurrence. This grants no new capability: the Site Owner can already stop production of receipts by powering the hardware off or refusing to host it (), and an explicit cessation assertion only makes that existing capability legible to a relying party instead of leaving it to be inferred from an absence of receipts.
• Retroactive impeachment MAY be asserted by the Issuer only. The Site Owner controls whether receipts are produced but not their content (), and an impeachment is an assertion about content that has already been produced and registered. Extending it to the Site Owner would grant a party with no authorship capability an authority over authored records that the trust model deliberately withholds.

Scope by attestation-binding mode The two classes apply in both attestation-binding modes of , and mean different things in each. An implementation MUST determine the mode before interpreting an assertion.

• In direct-witness mode, attestation.witnessKey matches iss, so both classes concern a single key and the Site Owner's cessation authority and the Issuer's impeachment authority attach to the same key material.
• In delegated-witness mode, attestation.witnessKey is distinct from iss. An assertion MUST identify the key it covers. An assertion covering the TEE signing key does not, by itself, assert anything about the Issuer's iss key, and an assertion covering iss does not, by itself, assert anything about the TEE signing key. A Verifier MUST NOT extend either to the other, and MUST NOT treat an assertion whose covered key cannot be determined as covering both.

Verifier behaviour Neither assertion deletes, invalidates, or suppresses a registered receipt. Registration is append-only and this document defines no mechanism by which a registered Signed Statement is withdrawn from a Transparency Service. A Verifier presented with a receipt for which it holds a relevant assertion:

• MUST surface the assertion to the relying party rather than resolving it internally;
• MUST identify which party made the assertion;
• MUST identify which of the two classes was asserted;
• MUST NOT suppress, discard, or downgrade the receipt on the basis of the assertion alone.

Neither assertion is self-authenticating, and this document does not adjudicate a disputed one. Where the Site Owner and the Issuer disagree, the profile supplies a detection property and not a remedy, in the same sense as . Adjudication is a matter for the relying party's own policy and for whatever legal or contractual regime governs the parties, and this document deliberately declines to make that determination on a relying party's behalf.

No payload member in this revision This revision defines no payload member carrying either assertion. An assertion about a witness key is a separate Signed Statement about a key, not a field inside a receipt about an engagement, and placing it in the receipt payload would require a receipt to be reissued in order to change a fact about its signer. Its content type and payload shape are deferred to a subsequent revision, and this revision states that they are deferred rather than reserving a member for them.

Revocation decision clock A receipt validly signed at time T whose witness key becomes subject to an assertion at a later time presents an ordering question, and the ordering MUST NOT be decided from a timestamp the signer supplied. ts, adapter.postedAt, and attestation.validity are all authored by the party whose key is in question, and a signer able to forge a signature is able to choose those values. Registration is mandatory in this profile (), so every conforming receipt carries at least one attached Receipt from a Transparency Service, obtained from a party other than the signer. A Verifier that orders a receipt against an assertion MUST derive the ordering from the registration of each, as evidenced by their attached Receipts, and MUST NOT derive it from any timestamp inside the receipt payload. This revision does not define an encoding for a Transparency Service's registration time, and does not require a Transparency Service to supply one. Where the Verifier cannot establish from the attached Receipts that one registration preceded the other, the ordering is undetermined, and the Verifier MUST surface it as undetermined rather than selecting an order. It MUST NOT fall back to a payload timestamp for this purpose, and MUST NOT substitute its own local clock. Stating this plainly is deliberate: a relying party writing policy against this profile needs to know that the profile carries the timebase requirement and does not yet carry the mechanism.

Three-party trust model The trust model described in this section applies to deployments where the TEE that produces receipts is physically hosted at the Site. In such deployments, the site owner both controls physical access to the TEE hardware and is the party responsible for its continued operation. This profile revision does not address deployments in which the TEE travels with a mobile Actor (for example, a TEE integrated into a mobile robot's compute platform), where the party controlling the attester's physical platform is distinct from the party controlling the Site. Such on-device attester topologies are not addressed here because a prerequisite is not yet in place, and naming that prerequisite is more useful than restating the deferral. A travelling TEE is a delegated-witness deployment: the platform is controlled by a party other than the Site Owner, so the authorization to sign on the TEE's behalf must be evaluated by a Verifier rather than assumed from physical custody of the hardware. That evaluation depends on resolving the delegation credential, which this revision defines as an Issuer-published fact () whose serialization is not yet specified. Until the encoding of that fact is fixed, a mobile-attester topology cannot be described in a way two implementations would evaluate identically, and specifying the topology first would produce a mode that reads as normative and cannot be conformed to. The security posture of this profile REQUIRES that three distinct parties participate in every receipt, and that no single party can produce a valid receipt alone:

• The site owner physically controls the TEE hardware. They can power it off, unplug it, or refuse to host it, but they CANNOT extract the signing key material or forge signatures with it. The site owner therefore controls whether receipts are produced at all, but not their content.
• The TEE silicon vendor provides the hardware root of trust that binds the signing key to a specific attested platform. Detection of a compromised or counterfeit TEE relies on this supply chain and is out of scope of this profile.
• The Issuer (typically the operator of a witness service) writes the Statement payload, causes the TEE to sign, registers the resulting Signed Statement with a Transparency Service, and performs the Adapter Write-In. The Issuer CANNOT sign without a live TEE. An Issuer that registers a receipt cannot prevent a relying party or auditor checking the Transparency Service from observing an equivocated chain. An Issuer that withholds a receipt from registration is not detected by this mechanism, which is why registration is mandatory in this profile ().

An implementation that collapses two or more of these roles into a single principal (for example, a cloud service that owns the TEE hardware AND signs AND registers with its own Transparency Service) is NOT conforming to this profile, and relying parties MUST NOT treat receipts from such an implementation as offering the trust properties defined here. Customer-controlled signing keys held outside a TEE are explicitly WEAKER than the model in this profile and MUST NOT be represented as equivalent. A site owner with direct access to the signing key can backdate, forge, or suppress receipts unilaterally, and no relying party -- insurer, regulator, or counterparty -- can distinguish an authentic receipt from a fabricated one in that setting.

Site Owner and Transparency Service as one principal This profile addresses two affiliation relationships. issuerAffiliation () states the relationship between the Issuer and the Site Owner. The disclosure of states the relationship between the Issuer and the Transparency Service. Neither states the relationship between the Site Owner and the Transparency Service, and this revision provides no member and no Issuer-published fact that carries it. The gap is not covered by the other two. An Issuer independent of both the Site Owner and the Transparency Service satisfies both existing disclosures, while a Site Owner that operates the Transparency Service the Issuer registers with still obtains, at the registration step, the ability to suppress or withhold entries concerning its own site. The external reference that requires is then external to the Issuer but not to the party whose conduct at the site is in question, which is the party a relying party is usually evaluating. A relying party that requires registration evidence external to the Site Owner cannot establish that property from a receipt conforming to this revision, and MUST obtain the relationship out of band. Stating this is deliberate: a policy author reading could otherwise conclude that an unaffiliated-Issuer registration establishes independence from the site, which it does not.

Identity attribution Identity attribution above the key level -- linking iss, actor.id, and site.id to real-world legal or natural persons -- requires an out-of-band identity binding document. This profile does not specify that document's format. -02 called it an identity manifest, which collided with the unrelated Issuer-published facts of ; the two were never the same document and the shared name implied they were.

Retained-content assurance limits The signed root binds recorded bytes and metadata; it does not authenticate the named asserting party or make the value true. A permitted category is not a credential. Raw evidence can be correctly hashed yet irrelevant, fabricated, stale, or contradictory. A relying party MUST NOT infer party authentication, key residence, physical presence, or actual time from membership, label equality, or raw-digest agreement alone. A valid subset cannot establish that hidden facts are well-formed, unique, or complete. A valid full block can still omit relevant events. Null can conceal an Issuer's choice not to commit available material, and a valid chain prefix can conceal a later suffix. These limits survive registration of the presented material; no complete-history witness is added here. Repeating a previously valid presentation does not make its signatures or commitments invalid. This revision's reference processing does not establish that an engagement is new merely because the same bytes verify again. Application decisions that require freshness or duplicate-use prevention need separately identified state, trust inputs, and policy. The local replay implementation in re-evaluates retained bytes and reports its findings; it is not a replay-prevention mechanism or a replay of the physical event. Salts provide no secrecy once exposed and no hiding guarantee when weak or predictable. Reused roots can link contexts. Consumers need bounded parsers, work limits, and access policies in addition to proof verification. A small inventory can be referenced repeatedly, so storage and hashing-work bounds are not interchangeable.

Position observations The recorded observation, its source attribution, and its preparation claims are not authenticated merely because the witness signed their commitment. Position spoofing, stale or fabricated samples, dishonest derivation labels, untrusted clocks, and substituted source identity remain possible. The implementation neither compares compound pose values with navigation evidence nor establishes hardware custody or party identity. Neither a pose nor its reference-point label changes the full retained envelope commitment. This addition does not implement a geometry engine, declare an older envelope's frame, evaluate containment, or turn an Issuer-asserted WITHIN into a recipient-verified conclusion. It does not add offline issuance, late-registration acceptance, blackout completeness, nested-site authority, or cross-witness independence.

Privacy Site identifiers, actor identifiers, and engagement types MAY be sensitive. Issuers SHOULD publish only the digests of envelope documents and internal evidence structures, as this profile requires. Issuers MAY additionally choose to encrypt the Statement payload under a per-relying-party key and publish only the Signed Statement's Receipt to a public Transparency Service, following the guidance in Section 6.2 for sensitive Statements. Position observations can expose sensitive location. Disclosing site.pose reveals its entire committed object at recorded precision; the construction does not redact individual coordinates or prove undisclosed values. Disclosures also reveal count, selected positions and ordering information, names, salts, and sibling hashes. A repeated root or evidence digest can enable linkage across statements. Retention and authorized access therefore need explicit privacy policies even when raw content is not public.

Implementation status This section follows . It describes implementation experience, not IETF endorsement, full conformance, or a requirement to use a particular implementation. It is to be removed before publication as an RFC. Revision -06 working snapshot. The source reviewed for this candidate is e725a53c2bbd90f3dcac1c06128ad4a367be5bd1. The historical integration and release observations below retain their original identities. This candidate claims no additional completed native-platform appraisal, CCF Receipt verification, or registered operational deployment. The digest-reference and evidence-exchange guidance in is design work; no completed cross-implementation result is attributed to it here. Source changes or experiments admitted after this snapshot require their own identified evidence before this section is updated. Reference implementation. pask-workspace, maintained by Wilder Robotics, is a Rust workspace containing pask-wire, pask-attest, pask-site, pask-adapter, pask-wire-cli, and pask-ts-client. Its source repository is https://github.com/wilder-robotics/pask-workspace. Public 0.7 source integration. At commit 60a3d16ed26841dc597b5b90f6fdb4ef5b039a5e (pull request 105, merged 2026-10-04 UTC), the public source implements the 0.7 retained-content construction, selective-disclosure verification, evidence-reference byte-integrity checks, the bounded unitless scalar comparator, recipient and replay checks, the site.pose vocabulary, and the version-specific text-string subject check. This identifies a reference-source snapshot, not a package release or a claim of full conformance. Initial 0.1.0 releases. Commit 9ba85f1034145de9e8670a83ffc26ca6608fcec6 corresponds to the initial 0.1.0 releases of the three permissive crates. That snapshot supports the 0.5 and 0.6 profiles. It does not implement the 0.7 retained-content construction, the additional vocabulary, or the subject-type correction specified by this document. The 0.7 source integration did not change those released artifacts or bump package versions. Package versions and profile versions are independent; the released 0.1.0 crates are not evidence of 0.7 support. Recorded checks. Pull-request workflow run 37174081694 checked the provisional test-merge commit 603f75dfee766cb110e3df4ece9e072378f5dcdf; post-merge workflow run 37178318604 checked the landed commit identified above. In both runs, the Cargo jobs passed formatting, lint, build, and default-feature and ES256-feature tests on Linux, macOS, and Windows. Separate Linux jobs performed document, revision, license, compliance, and name checks. The document-job passes are policy passes with recorded exceptions, not reports of zero idnits errors or IETF approval. The DCO and commit-name checks inspected the contribution range in the pull-request run; they were skipped on the post-merge push. The tracked-file name check ran in both. These results retain their respective execution identities. Released 0.2.0 crates. On 2026-10-12 UTC, pask-wire, pask-attest, and pask-wire-cli were published at version 0.2.0 on crates.io from public commit eddbb7f7c54d4e9aea59e8a0166d1c5165baf4d0 (pull request 108). The registry API and index checksums, and the downloaded archive bytes, matched the reviewed release artifacts. With the applicable features enabled, the released pask-wire exposes the 0.7 processing described above. The released CLI produces and verifies 0.7 Signed Statements from 0.7 input payloads using caller-supplied keys. Legacy 0.5 and 0.6 processing and the CLI's 0.5 canonical-example default are retained. These releases do not establish full profile conformance. Registry-release checks. Installation of pask-wire-cli 0.2.0 from crates.io and a separate registry-only consumer of pask-wire 0.2.0 exercised 0.7 production and verification with supplied Ed25519 keys, including rejection under an unrelated verification key. The CLI checks included a null contentDigest and a syntactically valid digest placeholder; the latter does not establish that committed content exists or was verified. Retained tests for pask-attest and pask-wire-cli passed from downloaded 0.2.0 archives using their own packaged lockfiles, with pask-wire resolved from the registry and without local path or Git overrides. These release checks used Cargo and rustc 1.98.1 on Linux x86-64. They are distinct from the source and document workflow executions recorded above, and do not constitute hardware-evidence appraisal or CCF Receipt verification. The initial 0.1.0 source snapshot, the first public 0.7 integration, and the 0.2.0 release source implement the payload and signature paths, recorded-time containment and DIRECT_WITNESS naming checks for 0.6, candidate-entry derivation, bounded Ed25519/RFC9162 Receipt validation, and sender byte-string wrapping. These components are not a claim of an end-to-end registered deployment. Service-key trust inputs, required claims, registration policy, and exact ledger representation require their own evaluation. CCF-profile Receipt verification is not implemented in any of these snapshots. The separate advisory TS Integration job exercises submission and Receipt retrieval, not Receipt-signature or inclusion-proof verification; its success is not evidence of CCF verification. The Signed Statement implementation in these snapshots checks that the Issuer identifier is nonempty text but does not fully enforce the inherited URI requirement in . This is an implementation limitation, not a relaxation of the normative requirement. The string-consistency check does not establish an authenticated association between that identifier and a signing key. None of these source snapshots, the registry-release checks, or this implementation-status report establishes authenticated asserting parties, navigation-source conversion, compound pose/evidence comparison, reliable real-world clocks, recipient geometry containment, hardware evidence appraisal, a complete mobile-site protocol, or latest/complete history. A disclosure-policy result must not be confused with application acceptance. In the published pask-attest 0.1.0 and 0.2.0 paths, verification of a quote containing evidence digests does not parse or appraise the referenced platform report and does not require those evidence bytes to be available. Those paths therefore cannot reject an otherwise verifiable quote merely because the referenced platform evidence is synthetic or absent. The 2026-09-29 UTC SEV-SNP experiment's narrative reports a separate AMD report and certificate-chain check. A sample-specific execution log for that check was not retained in the inventoried sample, so that narrative is not treated here as independently reproduced appraisal evidence. The 0.1.0 Pask crates used in that experiment verified the quote and Signed Statement, not the AMD report. The repository's known limitations record that configuration and its limits; the experiment does not establish hardware key custody for a deployed Issuer. The informative corrective-statement design remains unimplemented and its reference encodings remain unresolved. Planned experiments are not listed as completed interoperability results. The established repository split is Apache-2.0 for pask-wire, pask-attest, and pask-wire-cli, and AGPL-3.0-only for the operational crates.

Composition with external evidence records (informative) This section describes optional composition, not additional PSER wire requirements. It does not change profile wilder.pser/0.7, the existing payload or COSE-header rules, the content vocabularies, or the registration requirements in . It does not claim implementation of the external interaction or bundle profiles cited below. Those documents are individual Internet-Drafts and work in progress. An Agent Action Capsule can describe a software dispatch decision, while a PSER can record the subsequent physical-site engagement and attributed evidence. The records have different subjects and verification requirements. Shared text in a subject or correlation identifier is not authenticated identity, proof of causation, or proof that both records describe the same physical event. In particular, the 0.7 CWT sub remains text exactly equal to site.id; composition does not replace it with an action digest.

Preserve object and verification boundaries A composed presentation may contain a software-action record, admission material, an operator disposition, a PSER Signed Statement, attached SCITT Receipts, selected content disclosures, and a later review report. Each object retains its original bytes and its applicable profile. A signature or proof for one object is not a signature or proof for another. The software-action construction in is complementary to this profile. Generic COSE or SCITT checks can be shared where the exact byte construction, algorithms and trust inputs agree. They do not make an arbitrary action payload a PSER. A Capsule Producer Envelope is not substituted for this profile's Signed Statement or its protected claims. The named role of a key is evaluated under the relying party's selected policy. A conveyed key or certificate can participate in validation to independently accepted trust; it does not establish its own authority. Different keys alone do not demonstrate different organizations. A label asserting independence is not independent evidence of that relationship.

Digest contexts and original-byte identity A digest is interpretable only with the construction and byte boundary that produced it. The following contexts remain distinct:

Context	Input and intended conclusion
Original statement bytes	SHA-256 of the exact received COSE object, including its presented tag and unprotected header. Identifies that representation.
Complete canonical payload	An optional external reference to SHA-256 of the entire already-canonical JSON payload, including chain.hash. Identifies payload content, not a signer or transported envelope.
Payload chain.hash	The existing JSON-DIGEST after removing only nested chain.hash, preserving chain.seq, chain.prevHash, and every other payload member.
Retained-content root	The salted, domain-separated construction in , not JSON-DIGEST of the block document.
SCITT candidate entry	The exact derivation in , preserving the original protected, payload, and signature byte-string contents.

The complete-payload reference is described here as an optional composition context, not a new payload member, a replacement for id, or a new universal statement identifier. For this context, the field set is the complete 0.7 payload, the exclusion set is empty, the preimage has no added domain prefix, and the canonicalization is RFC 8785 JCS. In the local experiment described by this section, the comparison representation is sha256: followed by 64 lowercase hexadecimal digits. A carrier using raw hash octets or bare hexadecimal needs its own explicit conversion and comparison rule. For this context, the preimage is the exact attached payload byte string (element 2 of the decoded COSE_Sign1 array), after the receiver verifies the original statement and validates those bytes under the existing canonical- payload rules. The receiver hashes those preserved bytes directly, not a reserialized JSON object or a detached sidecar. A null or detached payload is unsupported by this local context. A noncanonical signed payload is rejected, not normalized to make this comparison succeed. Two signatures over the same payload can have the same complete-payload digest but different statement-byte digests and different signer associations. An application requiring a particular signed occurrence therefore identifies the signed object and its authenticated signer as well as its payload content. Canonicalization declarations in provide a vocabulary for describing such constructions. This revision does not select its optional header encoding, claim an artifact-type registration, or require CPB processing for PSER acceptance. Pask's nested chain.hash removal is not expressed by excluding the whole top-level chain object. A separate mapping needs to specify the nested transformation, its order, and textual representation. The retired jcs-n construction is not substituted for RFC 8785 JCS. A valid outer tag-18 representation and a locally accepted untagged representation can preserve the signature input while changing the digest of the entire presented object. Similarly, attaching a Receipt changes the transported object but preserves the candidate-entry derivation. These are properties of different byte domains, not license to rewrite retained originals or to infer replay equivalence.

Requests, answers, and coverage The interaction described by is useful for requesting a specifically identified PSER and associated evidence. A response can supply an artifact, provide an attributable refusal, or leave the requester with a record of nonresponse. This section does not specify a PSER request wire protocol or declare conformance to that interaction. For a bounded PSER experiment, a request identifies the exact statement representation or the explicitly selected payload context, the relevant issuer/site namespace, and the required evidence coverage. A content reference alone does not authorize a particular signed realization. The requester states whether it needs exact original statement bytes, an authenticated signed realization of identified payload content, or a selected disclosure. Coverage specifies the commitment or checkpoint being relied on and the properties requested from it. An inclusion Receipt for one statement does not authenticate a site's complete record range. Chain positions and signer-declared engagement times are not interchangeable with log positions or authenticated observation times. A request for an unsupported range, query, or freshness property is refused or reported unsupported, not answered as though it were a weaker successful request. A signed response needs to bind the exact request instance, the supplied artifact, and the selected coverage. Request-specific nonces, response times, and routing observations belong to the exchange record rather than being inserted into an otherwise deterministic artifact. Different disclosure selections are explicitly different requested views. Unauthorized content is not silently omitted while claiming to have supplied the requested view. A refusal signature supports attribution of the refusal to an authenticated responder. It does not prove that the responder's explanation is true. A timeout is a requester-observed interaction outcome, not a signature by the responder. Invalid response bytes and verification failures remain visible even where no usable terminal response was obtained. A signed retention undertaking states an obligation to keep specified material answerable until a stated time. Its signature does not demonstrate present possession, future availability, or disclosure authority. Evidence retention, content confidentiality, and a service's response policy remain separate concerns.

Absence, withholding, and physical activity The distinction between recorded claims and authenticated store properties in is useful when interpreting a PSER presentation. The scope of any negative conclusion is stated explicitly.

Observation or result	Supported scope
No usable response by the requester's deadline	The requester's recorded interaction, with its clock and transport limits.
Authenticated refusal asserting no such record	The responder made that assertion; record absence is not proved by the signature alone.
Missing sequence positions in a presentation	A discontinuity relative to the presented chain, not a measurement of missing physical work.
All committed slots disclosed and a supported name absent	Absence from that particular content block, under .
Separately verified index non-membership	Absence under the exact authenticated index, checkpoint, key/query semantics, and proof system.

This profile supplies no time-window non-membership construction and no proof that every physical engagement was captured. The existing chain and mandatory registration do not together supply that missing construction. Neither a site-resident TEE nor a neutral organizational role makes an unauthenticated search result a proof. An engagement not observed or not recorded can leave the same presented evidence as no engagement. A physical absence conclusion needs separate evidence about observation coverage, capture failures, time, and the relationship between events and committed records. Such a conclusion is not derived from successful PSER signature or chain verification alone.

Attribution and later review The original assertedBy, basis, and vocabulary-selected semantics are preserved. External evidence classifications can be described in a crosswalk only where their actual meanings agree; a label-to-label substitution is not sufficient.

Pask material	Interpretation boundary
appliance-measured with measured basis	The recorded provenance claim still needs evidence about what the appliance measured.
appliance-measured with estimated basis	An estimate is not relabelled as direct physical observation.
robot-attributed	The appliance may observe receipt of a robot message without observing the message's claimed physical quantity.
operator-entered	Human-entered material retains its attribution and authentication limits.
A checked signature, disclosure, or evidence digest	The identified check succeeded; this does not upgrade the original party or basis.

The verification findings in and the original provenance claims are separate dimensions. A later review can add findings about retained evidence, but it does not rewrite the committed record. An unsupported or contradictory fact does not become satisfactory because a later party signs a report containing it.

Verification reports and portable presentations The presentation discipline explored in is useful, but its Capsule-specific format is not adopted here. A bounded PSER report identifies original artifacts, the verifier implementation, applicable profile and policy, evaluation-time basis, trust inputs by identity, and each attempted check. For every finding, the report states the evidence used and whether the check passed, failed, was unsupported, remained unestablished, or was not evaluated, without allocating a new wire result enum. Integrity, signature validity, role authentication, inclusion, chain continuity, content disclosure, evidence integrity, native hardware appraisal, freshness, and application policy remain separate. Where a later party signs such a report, the signature needs to cover the findings and their stated scope as well as the artifact identities. Signing only a collection's digest while leaving its explanatory verdict mutable does not authenticate that verdict. Signature validity still does not establish that the reviewer is independent or authorized for the relying party's purpose. Citation closure means that the references required by a presentation are accounted for. It does not mean that all real events, all later statements, or all possible contradicting evidence have been supplied. A review report is not automatically a lifecycle assertion, amendment of an original PSER, or globally current verdict.

Admission and settlement composition A separately authenticated admission request and an operator disposition may be cited as retained supporting material. Their issuer authority, terms, scope and event ordering require their own checks. This section defines neither a bilateral control protocol nor a new witness-signature requirement. It does not change the meanings of DIRECT_WITNESS, DELEGATED_WITNESS, or the operations-layer acknowledgment in . The settlement-record model in provides one possible external consumer of engagement evidence. A reference to a PSER can identify evidence considered under agreed delivery terms; it does not by itself establish satisfactory work or payment entitlement. This profile does not actuate a robot, release payment, determine insurance coverage, or make a safety or compliance decision.

Interoperability claims An interoperability report identifies the exact implementations, inputs, keys and policies, byte constructions, algorithms, commands, outputs and negative controls. A shared-layer verification result is not PSER payload conformance, native hardware appraisal, third-party endorsement, or a production deployment. A peer's generic SCITT parser can authenticate a statement without understanding this profile's payload. Conversely, a PSER verifier is expected to refuse another application profile rather than reinterpret it as a PSER. Cross-implementation differences are reported before changing either profile's acceptance rules. Experiments that have not run are not included as successful implementation results.

References Normative References 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. International Association of Oil and Gas Producers Undated coordinate-reference-system definition. The resolver URI does not identify an EPSG dataset release date or version. The position representation in this document states its scaling and axis interpretation explicitly and does not select a WGS 84 realization or establish coordinate accuracy. Cryptographic operations like hashing and signing need the data to be expressed in an invariant format so that the operations are reliably repeatable. One way to address this is to create a canonical representation of the data. Canonicalization also permits data to be exchanged in its original form on the "wire" while cryptographic operations performed on the canonicalized counterpart of the data in the producer and consumer endpoints generate consistent results. This document describes the JSON Canonicalization Scheme (JCS). This specification defines how to create a canonical representation of JSON data by building on the strict serialization methods for JSON primitives defined by ECMAScript, constraining JSON data to the Internet JSON (I-JSON) subset, and by using deterministic property sorting. 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. 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. This document describes version 2.0 of the Certificate Transparency (CT) protocol for publicly logging the existence of Transport Layer Security (TLS) server certificates as they are issued or observed, in a manner that allows anyone to audit certification authority (CA) activity and notice the issuance of suspect certificates as well as to audit the certificate logs themselves. The intent is that eventually clients would refuse to honor certificates that do not appear in a log, effectively forcing CAs to add all issued certificates to the logs. This document obsoletes RFC 6962. It also specifies a new TLS extension that is used to send various CT log artifacts. Logs are network services that implement the protocol operations for submissions and queries that are defined in this document. This document describes how to include CBOR Web Token (CWT) claims in the header parameters of any CBOR Object Signing and Encryption (COSE) structure. This functionality helps to facilitate applications that wish to make use of CWT claims in encrypted COSE structures and/or COSE structures featuring detached signatures, while having some of those claims be available before decryption and/or without inspecting the detached payload. Another use case is using CWT claims with payloads that are not CWT Claims Sets, including payloads that are not CBOR at all. 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. 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. 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. 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. Informative References 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. In network protocol exchanges, it is often useful for one end of a communication to know whether the other end is in an intended operating state. This document provides an architectural overview of the entities involved that make such tests possible through the process of generating, conveying, and evaluating evidentiary Claims. It provides a model that is neutral toward processor architectures, the content of Claims, and protocols. NordenSoft Action State Group, Inc. Action State Group, Inc. Tyche Institute Action State Group, Inc. Action State Group, Inc. Action State Group, Inc. Independent Action State Group, Inc. 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. International Organization for Standardization Third edition. International Organization for Standardization Second edition. International Organization for Standardization First edition.

Corrective Statements (Informative) This appendix is informative and records a corrective-statement design for discussion. It does not define an interoperable corrective-statement wire format or impose additional requirements on 0.5, 0.6, or 0.7 engagement receipts. The corrective content type, versioning, and signed-statement reference encodings remain unresolved. No correction implementation or completeness of correction discovery is claimed. A correction is a SCITT Signed Statement registered with a Transparency Service. It is not a PSER engagement receipt. It carries its own content_type, distinct from the engagement-receipt content types, including application/pser+json; profile=wilder.pser/0.7. It does not carry the PSER spec member. It has its own unresolved versioning arrangement. The correction payload would carry:

• target (required, string): identifies the original signed assertion being corrected. Encoding is not frozen in this revision. The reference encoding is unresolved; no corrective reference format is adopted here. The reference mechanism must work for both engagement receipts and corrective statements.
• kind (required, string): the kind of corrective assertion. Admissible values:
  • AMENDMENT: the signer asserts that its own earlier statement was wrong, incomplete, or should no longer be relied upon. Same iss as the original is allowed. If a different iss is used, continuity or authorization must be established before the verifier reports it as that issuer's amendment.
  • CHALLENGE: the signer disputes another issuer's identified statement. Different iss is expected. Different issuer identifiers do not establish organizational independence or authority.
  • CORROBORATION: the signer supports a specified amendment or challenge. NOT labeled "independent support." Independence is reported separately when supported by evidence or policy.
• content (required, object): the corrective assertion. Would carry:
  • summary (required, string): human-readable description of what is being corrected.
  • fields (optional, array): specific payload fields of the original that are disputed or amended.
• ts (required, string): RFC 3339 UTC timestamp at which the correction was signed.
• amends (optional, object): identifies a specific earlier correction being withdrawn or replaced. Would carry:
  • ref (required, string): reference to the earlier correction. Same encoding as target.
  • operation (required, string): WITHDRAW or REPLACE.
  • Withdrawal and replacement authority is checked against the statement being changed (the earlier correction), not just against the original receipt. Both WITHDRAW and REPLACE operations require this check. A claimed withdrawal or replacement is attributable as an action by the earlier statement's issuer only when that identity or authorization relationship is established. Otherwise it remains another party's assertion about that statement.
• supports (required for CORROBORATION, string): reference to a specific amendment or challenge being corroborated. Same encoding as target. Every CORROBORATION identifies the specific statement it supports.

The signer declares the kind. The verifier reports which identity and relationship checks were established. The relying party assigns evidentiary weight. No universal different-iss requirement applies. Same-iss amendments are allowed. Different issuer identifiers do not establish organizational independence or authority. In this proposed model, a CHALLENGE would be evaluated against the public key of the statement being challenged. Where both relevant signatures have been successfully verified and a supported comparison establishes that the verification keys are the same, the verifier would report that the declared CHALLENGE fails the proposed distinct-key/kind-consistency condition. It would preserve the signed declared kind and report the reason separately. The statement's signature and any inclusion proof may still verify; the kind-consistency failure would not rewrite either signed record, silently convert the CHALLENGE to an AMENDMENT, or automatically invalidate the original. It would not determine the truth of the disputed facts. Kind consistency and its supporting key-relationship result would appear as explicit, machine-readable findings in the primary result for each evaluated corrective statement, and a failure or unestablished required relationship would be visible in the main report summary. They would not be available only as optional metadata. A broad summary that a correction is verified would not conceal a failed or unestablished required kind-consistency check. When a referenced statement, a successful signature verification, or the required comparison inputs are unavailable, the relationship would be reported as unestablished with a reason, not as different keys or as a successful consistency result. Issuer-signature, inclusion, kind-consistency, authority, original-statement, and local-policy findings would remain distinguishable. For CORROBORATION, the verifier would report the signing-key relationship to the statement named by supports and to the original anchor separately. Reuse of the supported statement's key would not count as an additional distinct-key source. Reuse of the original anchor's key alone would not force amendment classification. Different keys would not, by themselves, establish organizational independence. These paragraphs describe proposed reporting behavior. They do not finalize an output schema, field names, diagnostic-code registry, or corrective-statement wire format, and do not add a correction-processing requirement to wilder.pser/0.5, wilder.pser/0.6, or wilder.pser/0.7 engagement receipts. An illustrative reporting example, not a specified schema:

When O1 is unavailable, the key relationship and required kind consistency are unestablished with a reason. That is not portrayed as "different keys" or a successful challenge. These are proposed reporting examples, not claims that correction verification is implemented. The verifier evaluates presented statements and their direct references. It does not recursively discover corrections. It does not compute a "winning" history. Where processing limits or missing references prevent a relationship from being checked, the verifier reports that limitation rather than inventing a final disposition. Availability, verification, and search coverage are reported as distinct questions. No automatic winning history is computed. This design is presented as a proposal. Reference encoding for target, amends, and supports is not frozen and is coordinated with issue #67. The reference implementation does not produce or verify correction records. Discovery of corrections from an original receipt is not specified in this revision. "Original-only" means no correction was presented; it does not mean no correction exists. A correction does not reset the original's attestation.validity interval. A correction's ts is not constrained by the original's validity window.

Content vocabulary rule tables (normative) This appendix supplies every rule needed for the two closed vocabularies. The tables describe claims, not safety certification, evidence appraisal, or authenticated parties. The JSON blocks below use the finite dialect in . They are machine-readable rule definitions; they are not generic JSON Schema documents. Each row's meaning string is printed in the preceding Meaning: paragraph, and its other members appear in the JSON block. Restoring that one string reconstructs the complete row; display line wrapping inserts no new characters into the original string. The definitions and their prose are checked against the exact companion artifacts. For /2 use all 47 rows below. For /1 use the first 46 rows only, with all listed fields unchanged. The site.pose row is absent from /1. Global attribution, evidence, origin, and integer rules apply to both versions. The 46-fact vocabulary wilder.pser-content-vocab/1 has source-artifact SHA-256: 030cd709841fc057ba76f2568d44401b36d8ce823369756e11e2989309d4468d The 47-fact vocabulary wilder.pser-content-vocab/2 has source-artifact SHA-256: 2fb4e3a099003638d318333dee66fe2b710fb39b6dec78f570f6ecf31592a248 The committed digest prepends sha256: to the corresponding hash. The hash identifies the exact UTF-8 artifact, not the typography of this appendix. Implementations can reproduce the rule table from this appendix and recognize the named digest without retrieving an artifact at run time. The byte-exact JSON artifacts are companion implementation material. The complete rule definitions are also supplied here; an implementer does not need access to a private repository to interpret the two vocabularies. The artifact digests identify the unchanged source bytes, not a re-serialization of the displayed definitions. The /1 relation is 153 permitted and 813 forbidden among 966 known fact/party/basis combinations. The /2 relation is 157 permitted and 830 forbidden among 987 combinations. Counts are cross-checks, not substitutes for the complete row and global-rule intersection.

Global interpretation The seven party labels, three basis labels, global basis lists, and required evidence parties are defined in and . The following block contains the structured global metadata. Its three string annotations are printed immediately afterward as prose to keep the artwork within the text line limit. Reconstruct the complete object by restoring those strings under their named globalRules keys. This layout changes neither the source artifact nor its hash:

{ "assertedBy": [ "appliance-measured", "robot-attributed", "operator-entered", "platform-recorded", "site-policy", "manufacturer-declared", "undetermined" ], "basis": [ "measured", "estimated", "declared" ], "globalRules": { "basisByParty": { "appliance-measured": [ "measured", "estimated" ], "manufacturer-declared": [ "declared" ], "operator-entered": [ "estimated", "declared" ], "platform-recorded": [ "measured", "declared" ], "robot-attributed": [ "measured", "estimated", "declared" ], "site-policy": [ "declared" ], "undetermined": [ "estimated", "declared" ] }, "evidenceRequiredByParty": [ "appliance-measured", "robot-attributed" ] }, "remoteOrigin": [ "on-site", "off-site", "undetermined" ] }

Artifact annotation numbers: All numeric values are integers in the stated unit; floats are rejected (RFC 8785 number interoperability). Artifact annotation operatorPseudonym: crew.operator-pseudonym is assigned and mapped by the teleoperation platform; the witness never holds the mapping. Artifact annotation remoteOrigin: Tag values on-site | off-site | undetermined; required exactly where the table says, forbidden elsewhere. The artifact's parenthetical reference to RFC 8785 in its numbers annotation does not mean RFC 8785 prohibits all fractional numbers. Numeric fields in these fact schemas are integers; canonical commitment and raw evidence comparison retain their separately defined number models. The operator pseudonym is assigned and mapped by the teleoperation platform. The witness does not receive the identity mapping under this vocabulary. That custody condition is a producer/deployment obligation, not a result of the string-pattern validator. Similarly, statements in a row's meaning about a manufacturer, platform, or source's records remain claims whose truth and custody are not established by typed validation. remoteOrigin values and per-row requirement/forbiddance are interpreted as control-origin metadata, not generic evidence-acquisition metadata. No IANA registry is requested for these document-defined vocabulary labels or fact names. Extensions require a separately specified rule set and a new digest; a new label alone is not support or delegation of naming authority.

Complete fact rules

1. unit.model Meaning: Model designation of the acting machine.

{ "assertedBy": [ "manufacturer-declared", "robot-attributed", "site-policy" ], "basis": [ "declared" ], "evidence": "optional", "name": "unit.model", "remoteOrigin": "forbidden", "unit": null, "value": { "maxLength": 64, "type": "string" } }

2. unit.pseudonym Meaning: Stable pseudonym for the individual machine; the manufacturer holds the mapping to the physical serial.

{ "assertedBy": [ "manufacturer-declared", "robot-attributed" ], "basis": [ "declared" ], "evidence": "optional", "name": "unit.pseudonym", "remoteOrigin": "forbidden", "unit": null, "value": { "maxLength": 128, "pattern": "^[A-Za-z0-9:_.-]+$", "type": "string" } }

3. unit.deployment-status Meaning: Whether the site's registry lists this machine as a deployed unit or a planned one. Site-registry status, not proof of physical presence.

{ "assertedBy": [ "site-policy" ], "basis": [ "declared" ], "evidence": "optional", "name": "unit.deployment-status", "remoteOrigin": "forbidden", "unit": null, "value": { "enum": [ "deployed", "planned", "unknown" ], "type": "string" } }

4. unit.software-build Meaning: Identity of the software actually running (build string or measured digest).

{ "assertedBy": [ "robot-attributed", "manufacturer-declared" ], "basis": [ "measured", "declared" ], "evidence": "optional", "name": "unit.software-build", "remoteOrigin": "forbidden", "unit": null, "value": { "maxLength": 128, "type": "string" } }

5. unit.safety-parameter-digest Meaning: Digest over the active safety-function parameterization.

{ "assertedBy": [ "robot-attributed", "site-policy" ], "basis": [ "measured", "declared" ], "evidence": "optional", "name": "unit.safety-parameter-digest", "remoteOrigin": "forbidden", "unit": null, "value": { "pattern": "^sha256:[0-9a-f]{64}$", "type": "string" } }

6. unit.category Meaning: Category of the machine under the applicable safety standard; opaque code, checked for form only.

{ "assertedBy": [ "manufacturer-declared", "site-policy" ], "basis": [ "declared" ], "evidence": "optional", "name": "unit.category", "remoteOrigin": "forbidden", "unit": null, "value": { "maxLength": 32, "type": "string" } }

7. limits.rated-force Meaning: Rated force limit. A rated limit is a declaration; it is never measured.

{ "assertedBy": [ "manufacturer-declared", "robot-attributed" ], "basis": [ "declared" ], "evidence": "optional", "name": "limits.rated-force", "remoteOrigin": "forbidden", "unit": "N", "value": { "minimum": 0, "type": "integer" } }

8. limits.rated-speed Meaning: Rated speed limit. Declaration only.

{ "assertedBy": [ "manufacturer-declared", "robot-attributed" ], "basis": [ "declared" ], "evidence": "optional", "name": "limits.rated-speed", "remoteOrigin": "forbidden", "unit": "mm/s", "value": { "minimum": 0, "type": "integer" } }

9. unit.mass-centre-height Meaning: Height of the centre of mass, to bound fall energy.

{ "assertedBy": [ "manufacturer-declared", "robot-attributed" ], "basis": [ "declared", "estimated" ], "evidence": "optional", "name": "unit.mass-centre-height", "remoteOrigin": "forbidden", "unit": "mm", "value": { "minimum": 0, "type": "integer" } }

10. unit.total-mass Meaning: Total mass. Nameplate (declared) or weighed by site equipment (appliance-measured).

{ "assertedBy": [ "manufacturer-declared", "appliance-measured", "robot-attributed" ], "basis": [ "declared", "measured" ], "evidence": "optional", "name": "unit.total-mass", "pairsForbidden": [ [ "robot-attributed", "measured" ] ], "remoteOrigin": "forbidden", "unit": "g", "value": { "minimum": 0, "type": "integer" } }

11. unit.channel-rates Meaning: Declared native logging rate per safety channel, as a map of channel label to hertz.

{ "assertedBy": [ "robot-attributed", "manufacturer-declared" ], "basis": [ "declared" ], "evidence": "optional", "name": "unit.channel-rates", "remoteOrigin": "forbidden", "unit": "Hz", "value": { "type": "map-of-integer" } }

12. mode.control Meaning: Who or what was issuing motion commands. Carries remoteOrigin: whether the command source was off-site.

{ "assertedBy": [ "robot-attributed", "platform-recorded", "operator-entered" ], "basis": [ "measured", "declared" ], "evidence": "optional", "name": "mode.control", "remoteOrigin": "required", "unit": null, "value": { "enum": [ "autonomous", "supervised-autonomous", "assisted-remote", "full-remote", "fallback-to-remote", "transition" ], "type": "string" } }

13. mode.control-handover-preceding Meaning: Whether control changed hands within the 60 s before the event, and which way. Carries remoteOrigin.

{ "assertedBy": [ "robot-attributed", "platform-recorded" ], "basis": [ "measured", "declared" ], "evidence": "optional", "name": "mode.control-handover-preceding", "remoteOrigin": "required", "unit": null, "value": { "properties": { "direction": { "enum": [ "to-remote", "to-autonomy" ], "nullable": true, "type": "string" }, "occurred": { "type": "boolean" }, "secondsBefore": { "maximum": 60, "minimum": 0, "nullable": true, "type": "integer" } }, "required": [ "occurred", "direction", "secondsBefore" ], "type": "object" } }

14. scene.event-instant Meaning: Instant of the event at millisecond precision, on the receipt's clock basis. Measured by a device, never typed.

{ "assertedBy": [ "appliance-measured", "robot-attributed", "platform-recorded" ], "basis": [ "measured" ], "evidence": "required", "name": "scene.event-instant", "remoteOrigin": "forbidden", "unit": "ms UTC", "value": { "maxLength": 32, "pattern": "^[0-9]{4}-[0-9]{2}-[0-9]{2}T[0-9]{2}:[0-9]{2}:[0-9]{2}\\.[0-9]{3}Z$" , "type": "string" } }

15. scene.setting Meaning: The kind of place the work happened in, as the site declares it.

{ "assertedBy": [ "site-policy" ], "basis": [ "declared" ], "evidence": "optional", "name": "scene.setting", "remoteOrigin": "forbidden", "unit": null, "value": { "enum": [ "industrial-segregated", "industrial-shared", "logistics", "service", "domestic", "other" ], "type": "string" } }

16. scene.collaboration-regime Meaning: Collaborative operating regime in force; measured when the machine reports its active regime, declared when the site's plan states it.

{ "assertedBy": [ "robot-attributed", "site-policy" ], "basis": [ "measured", "declared" ], "evidence": "optional", "name": "scene.collaboration-regime", "remoteOrigin": "forbidden", "unit": null, "value": { "enum": [ "monitored-stop", "hand-guiding", "speed-separation", "force-limited", "non-collaborative", "transitioning", "unknown" ], "type": "string" } }

17. scene.persons-in-zone Meaning: Whether people were inside the working zone.

{ "assertedBy": [ "appliance-measured", "robot-attributed", "operator-entered" ], "basis": [ "measured", "estimated", "declared" ], "evidence": "optional", "name": "scene.persons-in-zone", "remoteOrigin": "forbidden", "unit": null, "value": { "type": "boolean" } }

18. scene.attendant-present Meaning: Whether an attending operator, local or remote, was present.

{ "assertedBy": [ "platform-recorded", "operator-entered", "site-policy" ], "basis": [ "measured", "declared" ], "evidence": "optional", "name": "scene.attendant-present", "remoteOrigin": "forbidden", "unit": null, "value": { "type": "boolean" } }

19. scene.surface-state Meaning: Condition of the floor or ground.

{ "assertedBy": [ "operator-entered", "appliance-measured", "site-policy" ], "basis": [ "estimated", "declared" ], "evidence": "optional", "name": "scene.surface-state", "remoteOrigin": "forbidden", "unit": null, "value": { "enum": [ "nominal", "wet", "uneven", "obstructed", "unknown" ], "type": "string" } }

20. event.kind Meaning: What kind of event this block describes.

{ "assertedBy": [ "appliance-measured", "robot-attributed", "operator-entered", "undetermined" ], "basis": [ "measured", "estimated", "declared" ], "evidence": "optional", "name": "event.kind", "remoteOrigin": "forbidden", "unit": null, "value": { "enum": [ "protective-stop", "emergency-stop", "transient-contact", "sustained-contact", "stability-loss", "fall", "near-miss", "hardware-fault", "software-fault", "sensing-fault", "override-attempt", "cyber-anomaly", "software-update", "remote-handover", "other" ], "type": "string" } }

21. event.trigger Meaning: What set the event off; free text bounded to 256 characters.

{ "assertedBy": [ "appliance-measured", "robot-attributed", "operator-entered", "platform-recorded", "site-policy", "manufacturer-declared", "undetermined" ], "basis": [ "measured", "estimated", "declared" ], "evidence": "optional", "name": "event.trigger", "remoteOrigin": "forbidden", "unit": null, "value": { "maxLength": 256, "type": "string" } }

22. event.trace-window Meaning: High-rate series across the mechanical event. The value describes the window; the series itself is the referenced retained object.

{ "assertedBy": [ "appliance-measured", "robot-attributed" ], "basis": [ "measured" ], "evidence": "required", "name": "event.trace-window", "remoteOrigin": "forbidden", "unit": null, "value": { "properties": { "durationMs": { "minimum": 0, "type": "integer" }, "rateHz": { "minimum": 1, "type": "integer" } }, "required": [ "durationMs", "rateHz" ], "type": "object" } }

23. event.scene-capture Meaning: Perception capture around the event. Frames are a retained object behind the reference; they never travel in the block.

{ "assertedBy": [ "appliance-measured", "robot-attributed" ], "basis": [ "measured" ], "evidence": "required", "name": "event.scene-capture", "remoteOrigin": "forbidden", "unit": null, "value": { "properties": { "biometricsExcluded": { "type": "boolean" }, "modality": { "enum": [ "video", "depth", "lidar", "mixed" ], "type": "string" }, "postMs": { "minimum": 0, "type": "integer" }, "preMs": { "minimum": 0, "type": "integer" } }, "required": [ "preMs", "postMs", "modality", "biometricsExcluded" ], "type": "object" } }

24. event.shift-trace Meaning: Low-rate series across the whole shift, to expose fatigue and slow drift. Series is the referenced retained object.

{ "assertedBy": [ "platform-recorded", "appliance-measured" ], "basis": [ "measured" ], "evidence": "required", "humanFactors": true, "name": "event.shift-trace", "remoteOrigin": "forbidden", "unit": null, "value": { "properties": { "durationMin": { "minimum": 0, "type": "integer" }, "rateHz": { "minimum": 1, "type": "integer" } }, "required": [ "durationMin", "rateHz" ], "type": "object" } }

25. event.override-attempt Meaning: Any attempt to override a safety function in the prior five minutes. Carries remoteOrigin as a separate tag.

{ "assertedBy": [ "robot-attributed", "platform-recorded", "appliance-measured" ], "basis": [ "measured", "declared" ], "evidence": "optional", "name": "event.override-attempt", "remoteOrigin": "required", "unit": null, "value": { "properties": { "attempted": { "type": "boolean" }, "secondsBefore": { "maximum": 300, "minimum": 0, "nullable": true, "type": "integer" } }, "required": [ "attempted", "secondsBefore" ], "type": "object" } }

26. event.software-update-preceding Meaning: Whether the machine's software was updated in the prior 72 hours.

{ "assertedBy": [ "robot-attributed", "manufacturer-declared", "site-policy" ], "basis": [ "measured", "declared" ], "evidence": "optional", "name": "event.software-update-preceding", "remoteOrigin": "forbidden", "unit": null, "value": { "properties": { "hoursBefore": { "maximum": 72, "minimum": 0, "nullable": true, "type": "integer" }, "occurred": { "type": "boolean" } }, "required": [ "occurred", "hoursBefore" ], "type": "object" } }

27. stability.recovery-engaged Meaning: Whether balance recovery engaged. Only the machine can know this.

{ "assertedBy": [ "robot-attributed" ], "basis": [ "measured" ], "evidence": "required", "name": "stability.recovery-engaged", "remoteOrigin": "forbidden", "unit": null, "value": { "type": "boolean" } }

28. stability.margin-min Meaning: Minimum stability margin reached (support-polygon margin). Only the machine can assert it; an operator cannot.

{ "assertedBy": [ "robot-attributed" ], "basis": [ "measured", "estimated" ], "evidence": "required", "name": "stability.margin-min", "remoteOrigin": "forbidden", "unit": "mm", "value": { "type": "integer" } }

29. stability.carried-mass Meaning: Mass being carried.

{ "assertedBy": [ "robot-attributed", "operator-entered", "site-policy" ], "basis": [ "measured", "declared", "estimated" ], "evidence": "optional", "name": "stability.carried-mass", "remoteOrigin": "forbidden", "unit": "g", "value": { "minimum": 0, "type": "integer" } }

30. stability.carried-speed-bound Meaning: Upper bound of load velocity.

{ "assertedBy": [ "robot-attributed", "site-policy" ], "basis": [ "measured", "declared" ], "evidence": "optional", "name": "stability.carried-speed-bound", "remoteOrigin": "forbidden", "unit": "mm/s", "value": { "minimum": 0, "type": "integer" } }

31. crew.session-minutes Meaning: How long the operator had been in session. Platform-recorded only; the operator cannot self-report it.

{ "assertedBy": [ "platform-recorded" ], "basis": [ "measured" ], "evidence": "optional", "humanFactors": true, "name": "crew.session-minutes", "remoteOrigin": "forbidden", "unit": "min", "value": { "minimum": 0, "type": "integer" } }

32. crew.supervision-ratio Meaning: Robots per operator at the time.

{ "assertedBy": [ "platform-recorded", "site-policy" ], "basis": [ "measured", "declared" ], "evidence": "optional", "humanFactors": true, "name": "crew.supervision-ratio", "remoteOrigin": "forbidden", "unit": "robots/operator", "value": { "minimum": 1, "type": "integer" } }

33. crew.certification-tier Meaning: Operator certification tier 1-3. Neither the operator nor the machine can assert it.

{ "assertedBy": [ "platform-recorded", "site-policy" ], "basis": [ "declared" ], "evidence": "optional", "humanFactors": true, "name": "crew.certification-tier", "remoteOrigin": "forbidden", "unit": null, "value": { "maximum": 3, "minimum": 1, "type": "integer" } }

34. crew.platform Meaning: Which teleoperation platform was in the loop.

{ "assertedBy": [ "platform-recorded", "site-policy" ], "basis": [ "declared" ], "evidence": "optional", "name": "crew.platform", "remoteOrigin": "forbidden", "unit": null, "value": { "maxLength": 128, "type": "string" } }

35. crew.operator-pseudonym Meaning: Stable pseudonym for the individual operator; the platform holds the mapping, the witness never does.

{ "assertedBy": [ "platform-recorded" ], "basis": [ "declared" ], "evidence": "optional", "humanFactors": true, "name": "crew.operator-pseudonym", "remoteOrigin": "forbidden", "unit": null, "value": { "maxLength": 128, "pattern": "^[A-Za-z0-9:_.-]+$", "type": "string" } }

36. link.rtt-mean Meaning: Mean round-trip latency over the prior 60 s.

{ "assertedBy": [ "platform-recorded", "robot-attributed" ], "basis": [ "measured" ], "evidence": "optional", "name": "link.rtt-mean", "remoteOrigin": "forbidden", "unit": "us", "value": { "minimum": 0, "type": "integer" } }

37. link.rtt-p95 Meaning: 95th-percentile round-trip latency over the prior 60 s.

{ "assertedBy": [ "platform-recorded", "robot-attributed" ], "basis": [ "measured" ], "evidence": "optional", "name": "link.rtt-p95", "remoteOrigin": "forbidden", "unit": "us", "value": { "minimum": 0, "type": "integer" } }

38. link.rtt-jitter Meaning: Round-trip latency variability over the prior 60 s.

{ "assertedBy": [ "platform-recorded", "robot-attributed" ], "basis": [ "measured" ], "evidence": "optional", "name": "link.rtt-jitter", "remoteOrigin": "forbidden", "unit": "us", "value": { "minimum": 0, "type": "integer" } }

39. link.loss-ratio Meaning: Packet loss over the prior 60 s, parts per million.

{ "assertedBy": [ "platform-recorded", "robot-attributed" ], "basis": [ "measured" ], "evidence": "optional", "name": "link.loss-ratio", "remoteOrigin": "forbidden", "unit": "ppm", "value": { "maximum": 1000000, "minimum": 0, "type": "integer" } }

40. contact.kind Meaning: Transient or sustained contact.

{ "assertedBy": [ "appliance-measured", "robot-attributed", "operator-entered" ], "basis": [ "measured", "estimated", "declared" ], "evidence": "optional", "name": "contact.kind", "remoteOrigin": "forbidden", "unit": null, "value": { "enum": [ "transient", "sustained" ], "type": "string" } }

41. contact.body-zone Meaning: Body region code under the applicable body model; opaque code, checked for form only.

{ "assertedBy": [ "operator-entered", "appliance-measured", "robot-attributed" ], "basis": [ "estimated", "declared" ], "evidence": "optional", "name": "contact.body-zone", "remoteOrigin": "forbidden", "unit": null, "value": { "maxLength": 32, "type": "string" } }

42. contact.force-peak Meaning: Peak contact force actually measured. Measured only. A reconstructed value can never be carried under this name.

{ "assertedBy": [ "appliance-measured", "robot-attributed" ], "basis": [ "measured" ], "evidence": "required", "name": "contact.force-peak", "remoteOrigin": "forbidden", "unit": "mN", "value": { "minimum": 0, "type": "integer" } }

43. contact.force-reconstructed Meaning: Contact force reconstructed where it was not measured. Estimated only, under a separate name.

{ "assertedBy": [ "appliance-measured", "robot-attributed", "operator-entered", "undetermined" ], "basis": [ "estimated" ], "evidence": "optional", "name": "contact.force-reconstructed", "remoteOrigin": "forbidden", "unit": "mN", "value": { "minimum": 0, "type": "integer" } }

44. contact.force-limit-active Meaning: Force-limiting threshold in force at the time.

{ "assertedBy": [ "robot-attributed", "site-policy", "manufacturer-declared" ], "basis": [ "measured", "declared" ], "evidence": "optional", "name": "contact.force-limit-active", "remoteOrigin": "forbidden", "unit": "mN", "value": { "minimum": 0, "type": "integer" } }

45. outcome.severity-grade Meaning: Severity 0 (near-miss) to 5 (fatality or permanent disability). A human judgment; no device can assert it.

{ "assertedBy": [ "operator-entered", "undetermined" ], "basis": [ "estimated", "declared" ], "evidence": "optional", "name": "outcome.severity-grade", "remoteOrigin": "forbidden", "unit": null, "value": { "maximum": 5, "minimum": 0, "type": "integer" } }

46. cause.category Meaning: Root-cause category. A judgment; no device, platform or site policy can assert it.

{ "assertedBy": [ "operator-entered", "undetermined" ], "basis": [ "estimated", "declared" ], "evidence": "optional", "name": "cause.category", "remoteOrigin": "forbidden", "unit": null, "value": { "enum": [ "hardware", "software", "sensing", "configuration", "balance", "local-operator", "remote-operator", "credential-compromise", "platform-failure", "network-quality", "environmental", "cyber-vulnerability", "cyber-supply-chain", "cyber-network", "unknown", "under-investigation" ], "type": "string" } }

47. site.pose Meaning: Position only: one observation of one site-declared reference point, from one source sample, scaled against EPSG:4979. Not orientation, speed, trajectory or ephemeris. With heightMm null it is a partial observation with no asserted height. The signature binds that this observation was committed with this attribution; it does not establish that the site was at this position, that the named source produced it, or its accuracy.

{ "assertedBy": [ "appliance-measured", "platform-recorded", "operator-entered" ], "basis": [ "measured", "estimated", "declared" ], "evidence": "optional", "name": "site.pose", "pairsForbidden": [ [ "platform-recorded", "declared" ], [ "operator-entered", "estimated" ] ], "remoteOrigin": "forbidden", "unit": null, "value": { "additionalProperties": false, "properties": { "crs": { "enum": [ "EPSG:4979" ], "type": "string" }, "heightMm": { "maximum": 100000000, "minimum": -20000000, "nullable": true, "type": "integer" }, "horizontalAccuracyMm": { "maximum": 100000000, "minimum": 0, "nullable": true, "type": "integer" }, "latE7": { "maximum": 900000000, "minimum": -900000000, "type": "integer" }, "latLonDerivation": { "enum": [ "exact-integer-rescaling", "rounded-half-even-to-1e-7-deg" ], "type": "string" }, "lonE7": { "maximum": 1799999999, "minimum": -1800000000, "type": "integer" }, "observedAt": { "maxLength": 32, "nullable": true, "pattern": "^[0-9]{4}-[0-9]{2}-[0-9]{2}T[0-9]{2}:[0-9]{2}:[0-9]{2}\\.[0-9]{3}Z$" , "type": "string" }, "referencePoint": { "maxLength": 64, "pattern": "^[A-Za-z0-9:_.-]+$", "type": "string" } }, "required": [ "crs", "referencePoint", "latE7", "lonE7", "heightMm", "horizontalAccuracyMm", "observedAt", "latLonDerivation" ], "type": "object" } }

Change log

Changes in -06 This revision preserves profile wilder.pser/0.7, its payload and header requirements, content construction, vocabulary artifacts, and legacy-profile processing rules. Internet-Draft, profile, and software package versions remain separate.

• Added informative composition guidance for software-action evidence, identified digest contexts, bounded evidence requests and review reports.
• Distinguished exact transported statement bytes, complete canonical payload content, chain.hash, retained-content roots, and registration candidate entries. No new universal statement identifier or CPB header requirement is introduced.
• Distinguished nonresponse, attributable refusal, presentation gaps, committed-block absence, and separately specified authenticated-index non-membership. No physical-time-window absence proof is claimed.
• Preserved original attribution when evidence is disclosed or reviewed. A later signature does not promote an assertion to a physical observation.
• Clarified that admission and settlement systems need their own authority, terms, and application decisions; neither control nor payment is added.
• Identified the current working source snapshot separately from published artifacts and historical executions. Narrowed the historical hardware experiment narrative to the evidence actually retained.
• Added five informative references pinned to specific revisions. The external drafts are not mandatory dependencies of PSER conformance.
• Qualified cross-record correlation so a shared subject does not imply same-event identity, and identified the complete-payload preimage as the original attached canonical bytes rather than a reserialized sidecar.
• Clarified that the informative corrective-statement proposal does not impose processing requirements on any supported engagement profile.

Changes in -05 This revision proposes profile wilder.pser/0.7 while preserving the declared meaning and signed bytes of the earlier profiles.

• Clarified that 0.5 and 0.6 reject engagement.contentDigest, even when null; the new member is confined to 0.7. The Issuer URI requirement is unchanged.
• Added required null-or-root engagement.contentDigest, the byte-exact retained-content construction, and selected/full disclosure verification.
• Defined the two closed, digest-selected vocabularies, their complete rule tables and finite value dialect. The seven recorded attribution categories are not authenticated party signatures.
• Separated raw-evidence integrity, supported value comparison, fact availability, and disclosure policy from membership and application acceptance; documented custody, null concealment, privacy, and suffix limits.
• Required the 0.7 protected CWT subject to be text exactly equal to site.id. The old implementation's byte-string compatibility path is preserved, not represented as CWT text conformance.
• Described the local portable recipient/chain workflow as implementation experience, not a new SCITT protocol or a physical-decision replay.
• Added optional site.pose only to vocabulary /2, without changing the operating-envelope commitment or adding public geometry. Producer conversion rules are distinct from the implemented typed-value checks.
• Added the CWT and EPSG reference needed for the new semantics. Corrective statements, hardware/key identity, offline issuance, blackout completeness, nested sites, and cross-witnessing remain outside this revision's additions.

Changes in -04 This revision introduces wilder.pser/0.6 as the profile version carrying the timestamp containment requirement and the DIRECT_WITNESS identifier-consistency convention. wilder.pser/0.5 retains its published meaning; its profile does not require timestamp containment or the naming convention. A receipt declaring wilder.pser/0.5 is not subject to these rules, regardless of when it was produced. This revision specifies the candidate-entry byte encoding for inclusion-proof verification, replacing the -03 disclosure that deferred this to an agreed convention. The candidate entry is the untagged four-element array [P, {}, M, S] derived from the presented Transparent Statement. Registration and verification alignment is required, not assumed. Transmitted envelopes use COSE tag 18; the untagged candidate entry is a profile-internal representation. A Transparency Service that hashes a different representation cannot satisfy this profile unless it commits to the candidate-entry representation specified here. The at-least-one-trusted-proof acceptance rule is specified: the profile's inclusion requirement is satisfied only when at least one attached Receipt has both a valid Transparency Service signature under an accepted service key and a valid inclusion proof for the derived candidate entry. A structurally parseable additional Receipt whose cryptographic verification fails does not defeat another attached Receipt that satisfies the requirement. The aggregate verification logic exists only as a test helper; application integration remains open. The TEE Class registry descriptions are clarified with informative context. No values are added, removed, or renamed. The evidence-format question for each value remains unresolved. This revision does not introduce an attestationResult member into the normative payload. The evidence-model question remains open. The complete explanation of retained evidence mechanisms, external appraisal, and the digest-versus-retrieval distinction is incorporated. Delegated signing does not itself establish evidence appraisal. This revision presents the corrective statement design as an informative appendix proposal. The correction payload type, signing rule, correction kinds, relationship model, processing rule, and verifier behavior are defined. Reference encoding for target, amends, and supports is not frozen and is coordinated with issue #67. The reference implementation does not produce or verify correction records. Every CORROBORATION identifies the specific statement it supports. Same-key CHALLENGE inconsistency is reported separately. Withdrawal and replacement authority is checked against the statement being changed. A correction does not reset the original's attestation.validity interval.

Changes in -03 This revision resolves a defect in -02 in which one undefined noun carried four unrelated obligations. -02 placed four normative requirements on a Verifier against "the Issuer's manifest" while stating, in its own identity attribution section, that it did not specify that document's format. A Verifier was therefore required four times to read a document the profile never described. The four obligations were not variants of one thing, and are not resolved by defining one document. It adds one REQUIRED payload member and bumps the profile identifier from wilder.pser/0.4 to wilder.pser/0.5. It removes nothing and narrows no existing requirement.

• attestation.bindingMode (REQUIRED) is added, carrying the attestation-binding mode in the receipt. In -02 a Verifier was required to obtain the mode from an Issuer-published document before it could rely on the non-extractability property stated in the introduction. The mode is a per-receipt fact fixed at signing time and known to the signer, and locating it outside the receipt made a property of the presented bytes depend on a network retrieval. The introduction's requirement now reads against the member. The value set is closed and a DIRECT_WITNESS receipt whose attestation.witnessKey and iss differ is rejected.
• The acknowledgement-object obligation is withdrawn as a Verifier requirement and restated as wording. -02 located the minimal ack object's schema in the Issuer's manifest in the same sentence that defined the object as bound by Merkle inclusion and not published. A document defined to be unfetchable cannot carry an obligation a Verifier can discharge. No mechanism changes; adapter.ackProvenance already distinguishes the case.
• is added, naming the two obligations that are genuine external retrievals and stating how they resolve. Both are disclosures about the Issuer rather than identity claims. The section fixes a stable Issuer-controlled identifier discoverable from iss, permits caching, voids a cached answer on a signing-key change, and defines a failure to resolve as undetermined.
• Undetermined is stated as a third outcome. A Verifier MUST NOT resolve an undetermined fact to the Issuer-favourable value, MUST NOT report a fact as absent where it was never successfully retrieved, and MUST NOT reject a receipt solely on the ground that a fact is undetermined. This follows the treatment already given to undetermined registration ordering in . The distinction between a fact checked and found absent and a fact never checked is load-bearing: collapsing the two lets a Verifier report to a relying party something it has no basis to state.
• No serialization format is specified for Issuer-published facts. This revision defines the obligations and their resolution behaviour and defers the encoding. Fixing a format before one has been deployed invites repudiation in the following revision rather than refinement.
• Conformance vectors accompanying this revision MUST include a failing resolution case for each Issuer-published fact, and the expected outcome of such a case MUST NOT be acceptance.
• The mobile-attester deferral in now names its prerequisite. -02 recorded that on-device attester topologies were expected to be addressed in a subsequent revision without stating what they waited on. A travelling TEE is a delegated-witness deployment, so it depends on resolving the delegation credential, whose encoding this revision deliberately leaves open.
• is added, recording that the profile carries no disclosure of the relationship between the Site Owner and the Transparency Service. An Issuer unaffiliated with both satisfies the two existing disclosures while a Site-Owner-operated Transparency Service retains the ability to withhold entries concerning its own site. The gap is recorded rather than closed.
• Editorial: the identity attribution section no longer calls its out-of-band document a manifest. It was never the same document as the Issuer-published facts above, and the shared name implied it was.

Changes in -02

• Corrected two internally inconsistent statements about the Site Owner that -00 and -01 both carried. The overview described the Site Owner as owning the hardware, while the trust model described the same party by capability; the overview now uses the capability language, matching the definition added to in this revision.
• Bounded the introduction's non-extractability statement to direct-witness mode. -01 stated as a profile-wide REQUIREMENT that the signing authority be neither extractable by the Site Owner nor by the Issuer, while normatively defining a delegated-witness mode in which the Issuer signs with its own key. The statement is now scoped to the mode it describes, the weaker property of the other mode is stated, and a Verifier is required to determine the mode from the manifest rather than assume it.

This revision closes the reviewer-identified gap in the Adapter Write-In, states a witness key lifecycle that -01 did not address, and records one scope boundary that -01 left to internal doctrine. It adds two payload members and bumps the profile identifier; it removes nothing.

• adapter.ackProvenance (REQUIRED) is added and the profile identifier and media-type parameter move from wilder.pser/0.3 to wilder.pser/0.4. -01 carried adapter.ackDigest with no way for a Verifier to tell an acknowledgement authored by an independent operations layer from one authored by the Issuer under the fallback in that member's own definition. The two are now distinguishable in the receipt rather than in out-of-band context. The member is REQUIRED rather than optional because an absent value would itself have to be assigned a meaning, which reintroduces the collapse.
• A Verifier MUST NOT resolve an unrecognized adapter.ackProvenance value in the receipt's favour. It is preserved as received and surfaced as unrecognized, and is neither read as THIRD_PARTY nor normalized to NONE.
• issuerAffiliation (REQUIRED) is added, stating whether the Issuer and the Site Owner are affiliated principals, with the three values AFFILIATED, INDEPENDENT and NOT_DISCLOSED. -01 gave a relying party no way to tell from a receipt whether the party that signed it had an interest in what it said, while requiring in that the roles not be collapsed. The member is REQUIRED for the same reason adapter.ackProvenance is: an absent value would have to be assigned a meaning, and every candidate meaning either manufactures a disclosure or makes an accusation.
• A Verifier MUST NOT read NOT_DISCLOSED as INDEPENDENT, and MUST NOT read an unrecognized value as AFFILIATED or normalize it to NOT_DISCLOSED. The profile records the Issuer's claim about itself and defines no mechanism for verifying it; a verified receipt is evidence that the claim was made and bound, not that it is true.
• A third Chain-Verifier obligation is added, and it is deliberately not a rejection. Where receipts presented as one chain carry different issuerAffiliation values, the change MUST be surfaced and identified by sequence number, and the presentation remains verifiable. Unlike chain.seq and chain.prevHash, which are wholly under the Issuer's control and admit no honest violation, affiliation is a relationship in the world outside the receipt and can legitimately change. What is prohibited is reducing a presentation to a single affiliation value, and in particular adopting the latest receipt's value, which is what would permit a chain to be relabelled after the fact by appending one receipt.
• Site Owner is added to , defined by capability rather than by title. -01 used the term in prose at three places in without defining it, and this revision is the first to assign it a normative capability.
• A witness key lifecycle is stated (), defining two named assertion classes. Cessation may be asserted by the Site Owner or the Issuer independently; retroactive impeachment may be asserted by the Issuer only. Neither deletes, invalidates, or suppresses a registered receipt. Both are scoped against the two attestation-binding modes of , and in delegated-witness mode an assertion MUST identify the key it covers.
• No payload member carries either assertion in this revision. The content type and payload shape of an assertion about a key are deferred, and this revision states the deferral rather than reserving a member.
• A revocation decision clock is stated (), requiring the ordering of a receipt against an assertion to be derived from registration rather than from any signer-supplied timestamp, and requiring an ordering that cannot be established to be surfaced as undetermined rather than guessed.
• Compromise of a TEE class or platform is stated not to be a key lifecycle event and remains out of scope ().
• A scope boundary is added to : the profile attests nothing about the internal state, intent, or decision process of a human participant, nor about signals conveyed by a direct neural or brain-computer interface. No member, value, or extension point is defined for one.
• Two implementation-status statements in -01 are corrected. The Chain-Verifier checks defined by -01 are now implemented, and the crate disagreement over a zero-length attestation validity interval is resolved with the rule stated in Section 4.1.

Changes in -01 This revision made two groups of changes. The first reconciles the profile identifier and four attestation members with the reference implementation and changes how the example figure in Section 4 is produced. The second corrects statements in -00 that were found to be wrong or unsupported, and adds normative requirements that -00 implied without stating. Both groups are enumerated below. Every normative change in this revision appears in one of them.

Reconciliation with the reference implementation

• The profile identifier and media-type parameter are wilder.pser/0.4. A producer built against wilder.pser/0.2 is rejected on version validation rather than on an unknown member.
• attestation.measuredBootChain (string) is replaced by attestation.measuredBoot, an object carrying the chain digest and the component sequence that hashes to it.
• attestation.platformEvidence is an object carrying a digest and an encoding label, rather than a bare string.
• attestation.validity is added and is REQUIRED. It carries notBefore and notAfter. notAfter MUST be strictly later than notBefore; a zero-length interval is rejected. This revision does not require a Verifier to test ts against the interval, and says so rather than implying a check that does not happen.
• The TEE Class registry is stated to be requested and not yet allocated, and the route by which a value is added is stated explicitly, so that an implementer on an unlisted confidential-compute environment has a documented path rather than only a rejection.
• The TEE Class registry values name confidential-compute environments rather than instruction set architectures.
• The Section 4 example is a complete, literal instance emitted by the reference implementation, and is asserted byte-identical to that implementation in its continuous integration. The -00 figure was a schema template rendered in a JSON code block and did not parse as JSON.

Corrections and added normative requirements

• The claim that the hash chain detects tail truncation is withdrawn. The chain field does not detect the withholding of the most recent receipts in any presentation, and does not detect equivocation. Section 7.2 is rewritten to state what the chain does and does not establish, and to attribute detection of either condition to evidence obtained from outside the presentation. The Abstract no longer asserts truncation detection, and the corresponding Section 1 scope bullet is rewritten. An appeal to TEE attestation as establishing recency is removed as unsound: a TEE establishes that it wrote the state it attests, not that that state is the most recent.
• The chain construction is now specified normatively in this document. -00 deferred part of it to ; that document is now cited for provenance only, and no conformance requirement of this profile depends on it.
• chain.seq is stated as non-negative rather than monotonic, and the first receipt in a chain MUST carry chain.seq 0. -00 used "monotonic", which does not constrain a single receipt and did not state the head value.
• Two chain-level verification requirements are added. A Chain-Verifier presented with two or more receipts as one contiguous chain MUST check chain.seq contiguity and MUST check that each chain.prevHash equals the preceding receipt's chain.hash. -00 described these properties as holding without requiring any party to check them. Chain-Verifier is defined in Section 2.
• chain.prevHash is redefined to remove an inconsistency that made the chain check unsatisfiable. -00 and an earlier draft of this revision defined chain.prevHash as a digest of "the immediately preceding receipt", while defining chain.hash as a digest taken with the chain.hash member absent. Read literally, those two definitions do not produce equal values, so the adjacent-pair check added above would have rejected every honest chain. chain.prevHash now carries the preceding receipt's chain.hash value by reference to that member rather than by an independent digest definition, and the exclusion is restated in both places. This was found by constructing a three-receipt chain and evaluating the requirement against it.
• Registration is now mandatory. An Issuer MUST register every receipt it issues with at least one Transparency Service, and a relying party MUST NOT accept a receipt as conforming without a verifying attached Receipt from a Transparency Service it trusts. -00 described registration as REQUIRED in its scope discussion without stating the requirement normatively. Where the Transparency Service is operated by the Issuer or an affiliate, that relationship MUST be disclosed and MUST NOT be treated as evidence external to the Issuer.
• A relying party SHOULD retain the highest verified chain.seq per chain as an anchor. The limits of that anchor, and the absence of any remedy defined by this profile, are stated explicitly.
• Section 7.5 no longer states that an Issuer cannot prevent an equivocated chain from being detected once registered. An Issuer that does not register is not detected by that mechanism; the mandatory-registration requirement is the response to that gap.

Acknowledgments The author thanks the SCITT WG for RFCs 9942 and 9943, and the authors of and for establishing the SCITT-AI receipt idiom on which this profile builds. The author thanks GitHub user giskard09 for a detailed public review of the -01 adapter and verifier semantics. That review identified that adapter.ackDigest was structurally unable to distinguish an acknowledgement authored by an independent operations layer from one authored by the Issuer, and pressed for the distinction to be carried in the receipt rather than left to out-of-band context. The adapter.ackProvenance member defined in is the result. The requirement that a value outside its closed set be surfaced as unrecognized, rather than read as THIRD_PARTY or normalized to NONE, is also his.