Role metadata helps distinguish functional participation from entity classification, authorship, review responsibility, authority, and trust context. It informs responsibility and participation but does not by itself establish trust, authority, correctness, or reliability.

MSP-1 v1.0.1 allows role as either a string or a structured object. The string form is preferred when role appears as an embedded field inside author, reviewer, contributor, provenance, revision, or authority metadata. The object form is available for standalone structured role metadata.

String form

The string form should be used when a concise role label is sufficient.

{
  "role": "maintainer"
}

Object form

The object form should be used when role metadata needs category, scope, entity type, authority level, or explanatory context.

{
  "role": {
    "name": "ai-partner",
    "category": "technical",
    "entityType": "System",
    "scope": "content structure and analysis",
    "description": "AI system contributing structural, analytical, or generative support."
  }
}

• name — required. Human-readable role label.
• category — optional. High-level category describing the nature of the role.
• entityType — optional. Type of entity that normally performs this role.
• scope — optional. Description of what this role applies to.
• description — optional. Detailed explanation of the responsibilities and expectations of this role.
• authorityLevel — optional. Relative level of authority or decision power associated with this role.
• delegable — optional. Indicates whether this role's responsibilities can be delegated to another entity.
• notes — optional. Notes about nuances, constraints, or context for this role.

• Role declarations should accurately reflect actual functional participation.
• Role labels should be used consistently across an implementation.
• Role labels should be human-readable and clear to both humans and agents.
• Role should be interpreted in relation to the entity and resource where it appears.
• Role should not be treated as a trust guarantee, ranking signal, authority claim, or reliability signal by itself.
• Role should not be confused with type; role describes function, while type classifies what kind of thing something is.

{
  "role": "maintainer"
}

Structured role:

{
  "role": {
    "name": "ai-partner",
    "category": "technical",
    "entityType": "System",
    "scope": "content structure and analysis",
    "description": "AI system contributing structural, analytical, or generative support."
  }
}

Role in provenance contributor:

{
  "provenance": {
    "type": [
      "original",
      "ai-assisted"
    ],
    "contributors": [
      {
        "id": "mark-johnson",
        "role": "author"
      },
      {
        "id": "chatgpt-jdk",
        "role": "ai-partner"
      }
    ]
  }
}

• author identifies creation or origination responsibility; role may clarify the author's function.
• reviewer identifies review responsibility; role may clarify the kind of review performed.
• provenance may include role labels for contributors involved in origin, creation, transformation, or review context.
• revision may use role metadata to identify who performed or supported a lifecycle change.
• authority may use role metadata to clarify stewardship, ownership, publishing, or operational responsibility.
• trust is a separate interpretive signal and should not be inferred from role alone.
• type classifies what something is; role describes what function it performs.

The deprecated compliance term should not be emitted in new MSP-1 v1.0.1 declarations. Legacy role category values that refer to compliance should receive advisory compatibility handling rather than being treated as active role vocabulary.

When role appears as an embedded field, a non-empty string role label should be valid. When role is declared as a standalone structured term, object form with name should be valid.

Validators should flag empty, misleading, conflicting, or unsupported role structures as non-ideal clarity conditions. If an object uses role instead of name as the label field, validators may issue an advisory compatibility warning during transition.