# Dataset field dictionary

Version reviewed: 2026-08-27.

This dictionary describes the four public exports listed in the [dataset manifest](https://agenticcommerceprotocol.info/dataset-manifest.json). The public copy of this dictionary is available at <https://agenticcommerceprotocol.info/dataset-field-dictionary.md>. A record documents what a cited source supported at a dated review. It is not a protocol certification, merchant deployment test, market-share estimate or independent confirmation of production adoption.

No separate reuse license or DOI is asserted. Download availability and a checksum do not grant reuse rights.

## Evidence registry

Public files: `evidence.csv` and `evidence.json`.

The JSON export contains the following top-level fields:

| Field | Type | Definition |
|---|---|---|
| `dataset` | string | Human-readable dataset name. |
| `datasetVersion` | ISO date string | Latest `last_verified` value present in the exported evidence records. |
| `methodology` | URL string | Public methodology used to review and classify evidence. |
| `evidence` | array | Evidence records described below. |

Each evidence record and CSV row contains:

| Field | Type | Definition |
|---|---|---|
| `claim_id` | string | Stable local identifier in the form `EV-###`. |
| `page_or_component` | string | Site page, dataset section or component that uses the claim. |
| `claim` | string | Bounded statement reviewed against the cited source. |
| `source_url` | URL string | Source inspected for the claim. |
| `source_type` | string | Editorial classification of the source, such as `official`. |
| `date_accessed` | ISO date string | Date the source was accessed for the evidence record. |
| `last_verified` | ISO date string | Most recent date the claim and source relationship was reviewed. |
| `evidence_excerpt` | string | Short evidence note or excerpt used during verification. It is context, not a replacement for the source. |
| `confidence` | string | Editorial confidence label recorded by the project. It is not a statistical probability. |
| `status` | string | Current editorial verification state of the record. |

## Protocol conformance inventory

Public files: `conformance.csv` and `conformance.json`.

The JSON export contains:

| Field | Type | Definition |
|---|---|---|
| `dataset` | string | Human-readable dataset name. |
| `datasetVersion` | ISO date string | Review date shared by the exported inventory. |
| `methodology` | URL string | Public methodology URL. |
| `interpretation` | string | Rules for interpreting `published`, `not-identified`, validators and TCKs. |
| `reuseStatus` | string | Explicit notice that no separate reuse license has been published. |
| `records` | array | Protocol artifact records described below. |

Each protocol record and CSV row contains:

| Field | Type | Definition |
|---|---|---|
| `protocol` | string | Protocol or proprietary program name. |
| `slug` | string | Stable site identifier. |
| `reviewed` | ISO date string | Date on which the listed official sources were reviewed. |
| `release` | string | Release label observed or a bounded explanation when no tagged release is asserted. |
| `releaseUrl` | URL string | Official page supporting the release label. |
| `specification` | enum | `published`, `not-identified` or `not-applicable`. |
| `specificationUrl` | URL string or empty string | Official specification location when identified. |
| `machineReadable` | enum | Status of a public normative machine-readable artifact. |
| `machineReadableUrl` | URL string or empty string | Artifact location when identified. |
| `conformanceSuite` | enum | Status of an official public conformance suite or TCK. |
| `conformanceUrl` | URL string or empty string | Suite or TCK location when identified. |
| `validationTool` | enum | Status of a distinct public validation tool. |
| `validationToolUrl` | URL string or empty string | Validator location when identified. |
| `limitation` | string | Scope boundary and facts that were not established by the review. |

`not-identified` is limited to the reviewed sources and review date. It is not proof that an artifact cannot exist elsewhere or later.

## Integrity manifest

Repository file: `docs/DATASET_MANIFEST.json`. Public URL: <https://agenticcommerceprotocol.info/dataset-manifest.json>.

| Field | Type | Definition |
|---|---|---|
| `manifestSchemaVersion` | string | Version of this manifest structure. |
| `releaseVersion` | ISO date string | Latest dataset version among the included artifacts. |
| `manifestUrl` | URL string | Stable public URL of the generated manifest. |
| `methodology` | URL string | Public methodology URL. |
| `fieldDictionary` | URL string | Stable public URL of this dictionary. |
| `licenseStatus` | string | `not-specified` until an explicit license decision is reviewed and published. |
| `license` | null or string | `null` in this release because no license is asserted. |
| `doi` | null or string | `null` until a real archive deposit is published. |
| `integrityAlgorithm` | string | Hash algorithm, currently SHA-256. |
| `integrityScope` | string | Bytes covered by each digest. |
| `artifacts` | array | Path, public URL, media type, dataset version, record count, byte length and SHA-256 for every export. |

## Reproduce the checksums

From a clean checkout with the locked dependencies installed:

```sh
npm run build
node scripts/build-dataset-manifest.mjs --check
```

To refresh a manifest deliberately after a reviewed dataset change:

```sh
npm run build
node scripts/build-dataset-manifest.mjs
node scripts/build-dataset-manifest.mjs --check
```

The script hashes exact files from `dist`. A different build output must produce a different digest and fail `--check` until the manifest is intentionally regenerated.