Claims and references
A party may rely on a policy, a lived experience, a professional standard, a dataset, or a community account. The ledger records the source and its intended role instead of calling every input “fact”.
An agreement is more than a signature. It can include claims, references, responsibilities, conditions, and different understandings of the people involved. An agreement ledger gives those parts a visible home before anyone asks software to act on them.
The result is not a system that decides what people “really meant”. It is a record of what each person put forward, what they accepted, what remains open, and the exact rule set they jointly authorise for a defined purpose.
A four-part path
Keep propositions, references, assumptions, and conditions distinct. A recorded claim is not automatically treated as an established fact.
Each party can state what they accept, question, qualify, or reject — including the scope and confidence of their understanding.
The shared agreement contains only accepted terms. Dissent, exceptions, and unresolved points remain visible instead of being silently flattened.
The agreed terms become a Semantic Instrument with declared inputs, limits, version, parties, and receipts for later runs.
Different understandings are data
A party may rely on a policy, a lived experience, a professional standard, a dataset, or a community account. The ledger records the source and its intended role instead of calling every input “fact”.
A person can say “I accept this for this situation”, “I need more evidence”, or “I agree only with this exception”. These qualifiers help prevent a broad rule being inferred from narrow consent.
A disagreement is not a broken record. It can be retained with the agreement, revisited when conditions change, and reflected in a later version without rewriting history.
Illustrative example
A resident group, a local service, and a community worker want a consistent way to prepare referrals. They do not hand authority to the software; they write a narrow shared procedure for handling a referral package.
| Recorded material | Party perspectives | Consolidated term |
|---|---|---|
| Claim: referrals need enough information to be reviewed. | Service: identity and contact fields are needed. Resident group: people must be told what is missing. | Require the declared referral fields; return a clear missing-information status. |
| Reference: local service criteria and the group's consent guidance. | Worker: a person must review sensitive cases. Service: criteria can change by version. | Include the criteria version and route sensitive or uncertain cases to human review. |
| Boundary: a prepared referral is not a decision on eligibility. | All parties accept this limit. | Never output “eligible”; use ready, held, or needs-human-review. |
When an agreement takes effect
Writing a rule is not the same as authorising it. The agreement lifecycle makes that distinction visible: a software agent may use an instrument only while the relevant human agreement is current, within scope, and supplied with the declared information.
Claims, references, terms, and party positions are visible. No agent authority exists yet.
The named parties meet the declared consent condition and approve the specific version.
The bounded instrument can follow the ratified procedure and retain a receipt.
Missing inputs, an unmet condition, or an unresolved exception hold the run rather than invent an outcome.
A revised or withdrawn agreement stops new reliance; earlier receipts retain their context.
Temporal provenance and review
A human-centric instrument records the evidence, assumptions, authority, and limits that were available at the time of a decision. Later evidence does not rewrite that original context. It creates a new, time-linked record that can trigger review, revision, pause, relief, or remedy for ongoing negative effects.
Capture the versioned evidence, known uncertainties, mandate, and reason the decision was made.
Link new material, changed conditions, affected parties, and the date it became available without overwriting T₀.
Apply the declared review trigger: confirm, qualify, hold, revise, or stop further reliance on the earlier rule.
Record relief, correction, escalation, or a revised agreement—and retain the full chain for later review.
Optional integrity anchoring
An agreement, provenance record, review event, or receipt may remain in the relevant local or controlled environment. When the parties choose, it can also carry an integrity commitment: a content digest, a Merkle-tree inclusion proof, or a reference pinned to a distributed ledger (DLT). Later records can link to the earlier commitment and to the revised version.
An anchor can help show that a particular commitment existed and has not changed relative to that chain. It does not itself prove the underlying statement is true, that a person consented, or that a record has a particular legal effect. Sensitive material and unnecessary personal references should not be placed in a public ledger.
Zero-knowledge proof support
Where a suitable proof circuit exists, QualiaDB/Webizen can use zero-knowledge proofs to show that a defined predicate or policy condition has been satisfied without exposing the private witness material. This lets an agreement or instrument ask the narrow question it needs answered while retaining privacy over supporting records.
Prove a threshold, range, authorised condition, or other supported predicate without disclosing the private value, source record, or full credential.
The verifier checks the specific circuit, public inputs, and proof. The result says only that the declared condition was satisfied for that circuit.
A receipt can record the proof type, policy or circuit version, verification result, permitted purpose, and time—without publishing the private witness.
Disclosure is not binary
Zero knowledge means that verifying the proof does not reveal its private witness. It does not mean that the underlying record is metaphysically unknowable to every other person or system. A protected-person, witness-protection, safeguarding, or similar instrument may deliberately separate a public or routine proof from controlled custody of the supporting evidence. An authorised court or other accountable decision-maker may be able to request or receive the necessary material under the applicable process, without making it available to routine verifiers, software agents, or the public.
Receives the declared statement, public inputs, proof, and a scoped verification receipt. It need not receive names, locations, source records, or other witness material.
Holds the sealed underlying material under a declared custody and access policy. Holding encrypted material does not itself confer permission to inspect, copy, or disclose it.
May receive the least material necessary only when an accountable human authority, stated purpose, time limit, and applicable procedure permit it—for example, a court acting through the relevant legal process.
Identifiers and evidentiary basis
An instrument may contain one or more scoped identifiers connected with a person, organisation, licence, qualification, role, device, or permission. For example, it may refer to a professional licence, a board-approved role, a course credential, or a device permission such as an API key. These references make the relevant records linkable and checkable; they do not define a human being or create authority on their own.
This is why instruments can support advanced software and AI systems: they provide structured, attributable material for a limited task while preserving the basis on which the task was performed.
Layer 4 · Formal Specification · Identifier Fabric (IDF-01–IDF-05)
Agreement records are validated against ontologies/identifier_fabric_shapes.ttl across four disjoint planes:
Entity (Who Plane)
NaturalAgent (living-SHACL), LegalEntity, and SystemAgent. Never merged by similarity, handle, wallet, location, or biometrics; owl:sameAs is forbidden on the Who plane (SameAsError::WhoPlaneForbidden).
Claim Plane
Scoped, signed assertions (VerifiableCredential, idf:ClaimOpinion, idf:RoleCapacityGrant, deontic/epistemic norms) with provenance and expiry. VCs prove cryptographic origin integrity of an issuer's statement, not universal ontological truth.
Handle Plane
Revocable, rotatable network or directory locators (DNI / idf:DigitalNetworkIdentifier, idf:RelationScopedLocator, DNS, email, did:qi endpoints). Revoking a handle closes the channel without erasing the Entity.
Instrument Plane
Cryptographic keys, hardware tokens, and software tools (idf:Instrument, idf:AiAgent, idf:CoAttestationBundle) classified by InstrumentProofClass (CryptographicOriginIntegrity, HardwareChannelPossession, SemanticClaimEvaluation) and closed KeyRole (IdentityAssertion, CredentialSigning, TransportHandshake, StorageEncryption, RecoveryBootstrap).
Cross-plane links use typed evidence morphisms (idf:supportedByInstrument, idf:assertedAbout, idf:observedAtHandle), while the F5 (compile-time VibeScript) and F6 (runtime ingress/query) Anti-Feature Firewalls reject person_score, identity_confidence, and same_person.
Structural rule · who acted, and in what capacity
This is a required property of a human-centric system, not an optional provenance feature. Where an agreement or instrument involves an incorporated entity, it must preserve the accountable human decision-makers or authorised representatives, the role in which they acted, and the scope of that authority. The point is accountable provenance, not indiscriminate disclosure of personal information.
The instrument must distinguish the incorporated entity from the people who made or authorised the relevant decision. Its provenance records:
An AI agent may carry out a bounded task for a natural person acting personally, or for a natural person exercising an authorised role for an incorporated entity. It is never the source of authority itself. The receipt must record:
Provenance as a Semantic Instrument
A provenance instrument is a bounded package that records the origin and development of a work, decision, dataset, agreement, or technical artefact. Other Semantic Instruments can refer to it as evidence: not merely that a record exists, but how it was made, who contributed, which authority applied, and what terms governed its use.
Versions, source material, technical artefacts, dates, contribution events, reviews, and the relationship between them.
The people and roles involved in creation, approval, assignment, stewardship, or later authorised use—within an appropriate privacy scope.
Applicable agreements, licences, assignments, confidentiality conditions, and the limits on what an agent or recipient may do.
Community of Experts · proposed instrument profile
Advanced research in a field may produce models, datasets, methods, findings, visualisations, and supporting source material. A Community of Experts (COE) Semantic Instrument can record the field-specific community’s scope of review and the conditions under which those artefacts may be relied upon. It is a credential about a defined community process—not a claim that a model is universally correct or that an expert community replaces human responsibility.
The Hyper Media Container carries the inspectable material: a description, definition graph, manifest, artefacts, model documentation, references, and integrity information. It can include compact semantic payloads while declaring larger models or data volumes as dependencies.
An HMC is transparent and integrity-checkable. Confidentiality or restricted access requires separate, declared protection and access conditions; it is not implied simply by putting material in a container.
The companion Semantic Instrument states what the expert community reviewed, the applicable field and purpose, the people and roles authorised to attest, the evidence and dissent considered, and the limits of permitted reliance by people or software agents.
Its receipt can bind a later use to the research-package version, review status, rights conditions, declared model limits, and the human review or escalation path.
| COE record | What it establishes within scope | What it does not establish |
|---|---|---|
| Research package | Which artefacts, versions, sources, and model dependencies were supplied. | That every result is correct, safe, or suitable for every purpose. |
| Community attestation | Which defined review, contribution, or approval process occurred and under what terms. | That the community has unlimited authority or that its view is beyond challenge. |
| Use contract | The authorised task, licence conditions, declared limits, and review route for a person or AI agent. | Permission for an agent to extend its task, disclose protected material, or substitute for required human judgement. |
When it becomes an instrument
Important boundary
People remain responsible for deciding whether to adopt an agreement, whether its sources are appropriate, and when it should be reviewed. For a closer look at the records that make this possible, continue to the agreement ledger.
Explore the agreement ledger