# Human/agent documentation parity worksheet

Blank review record. It is not evidence that a documentation build ran, either surface exists, an agent fetched or understood content, accessibility was evaluated, or publication was approved.

## Review identity

- Review ID:
- Reviewed source revision/hash:
- Source owner:
- Human surface URL:
- Agent-readable surface URL:
- Agent index URL, if applicable:
- Review scope:
- Review status: Not reviewed

## Surface inventory

- Human surface present: Not reviewed
- Agent surface present: Not reviewed
- Human navigation/discovery path:
- Agent discovery path:
- Authentication/access boundary:
- Build or generation evidence reference:
- Publication evidence reference:

## Claim ledger

Add one row per required meaning. Blank cells and `Not reviewed` never mean parity.

| Claim ID | Source meaning | Required exact literals/targets | Human surface evidence | Agent surface evidence | Allowed difference | Review result |
| --- | --- | --- | --- | --- | --- | --- |
|  |  |  |  |  |  | Not reviewed |

Allowed review results: `Present`, `Missing`, `Contradictory`, or `Not reviewed`.

## Cross-surface checks

- Required warning parity: Not reviewed
- Scope/non-goal parity: Not reviewed
- Configuration literal parity: Not reviewed
- Link-target parity: Not reviewed
- Version qualification parity: Not reviewed
- Recovery-path parity: Not reviewed
- Code example parity: Not reviewed
- Known intentional omissions:
- Undeclared differences:

## Separate quality reviews

- Human accessibility review: Not evaluated
- Agent formatting/size review: Not evaluated
- Authentication/authorization review: Not evaluated
- Search/index discoverability review: Not evaluated
- Agent retrieval observed: Not evaluated
- Agent comprehension evaluated: Not evaluated

## Verdict

- Reviewer:
- Review time:
- Parity verdict: Unresolved
- Supported claim, stated narrowly:
- Exceptions and limitations:
- Required correction:
- Correction owner:
- Re-review required: Yes

Allowed parity verdicts: `Supported`, `Supported with exceptions`, `Unresolved`, or `Failed`. Discoverability is not authorization, matching structure is not semantic parity, and a blank evidence field is never a pass.