# Verifier behaviour

**Design guidance.** This page describes the proposed behaviour of a verifier for Open Asset Evidence. It does not establish that a particular service implements these checks or is ready for production. Read the [assurance and release status](assurance-and-release.md) for that distinction.

A signature check establishes that signed content matches a signature under the verification key. Identifying the issuer, establishing its authority and deciding whether to use the evidence require additional checks. A signature does not establish that an observation or valuation is factually correct.

## Inputs the recipient controls

The proposed verifier separates three input groups:

- **Presented evidence**: Source: Holder or presenting party. Purpose: Credentials, disclosed claims and supporting references to check.
- **Trust and status information**: Source: Sources accepted under the recipient's policy. Purpose: Authenticated issuer, key, authority and status information, with its scope and freshness.
- **Recipient context**: Source: The recipient. Purpose: Expected asset, intended purpose, policy version, acceptable freshness and presentation context.

Evidence supplied by a presenter must not weaken the recipient's requirements. A public key or address embedded in a credential is not, by itself, an authenticated link to an issuer. Key resolution needs the recipient's accepted [trust inputs](../shared/policy/trust-list.md).

## A result with separate checks

The proposed [Verification Result Contract](../shared/verification-result-contract.md) separates the following questions:

- **`L1_parse`**: Can this format, type and version be interpreted?
- **`L2_signature`**: Are the signature and issuer identity established, and are applicable presentation-binding checks satisfied?
- **`L3_authority`**: Is the issuer recognised for this statement and scope?
- **`L4_status`**: Are validity and credential status acceptable at the time of checking?
- **`L5_semantics`**: Does the evidence refer to the expected asset, engagement and related records?
- **`L6_policy`**: Does the presented evidence satisfy the recipient's purpose and requirements?
- **`L7_acceptance`**: Has the recipient recorded its own decision?

Technical `ALLOW` requires the first six layers to be `PASS`. Missing, stale, unavailable and unsupported checks retain their stated outcomes. A required check that cannot be completed prevents technical `ALLOW`; insufficient evidence is distinct from an invalid signature or a contradiction. Other result producers leave the recipient's acceptance `NOT_ASSESSED`.

A result should identify the evidence presented, policy used, time checked and limits on reuse. It should expose the individual checks rather than rely on a single unqualified “verified” label. The current reference result checker checks the consistency of a reported result; it does not verify credential signatures or demonstrate a complete verifier.

## Freshness and disclosure

Offline verification depends on the trust and status information available locally. A cached record supports a currentness check only within the recipient's accepted freshness limits. An unavailable status source is different from a revoked credential, but it does not supply a passing status check. A presenter cannot extend an acceptable age by declaring a longer limit.

[Selective disclosure](../shared/evidence-presentation-and-disclosure.md) limits what is shared. Withholding a required claim may prevent a check from completing. A valuation and a separately signed condition record each need their own checks; a reference between them does not authenticate either record by itself.

Technical verification does not grant legal reliance, professional indemnity cover or institutional acceptance. Those boundaries are explained in [Trust and responsibilities](trust-and-responsibilities.md).
