Skip to content

delegation.md draft + AAE / W3C VC alignment — observations and offer #29

@MoltyCel

Description

@MoltyCel

Hi @uchibeke,

Read through OAP v1.0, the delegation.md working draft, and the arXiv paper. Your "complementary, not competing" framing in the conclusion is the right characterization, and it matches what I see when mapping OAP against W3C VC + DID architectures.

A few structural observations, plus one concrete offer.

Structural convergence in delegation.md

The v1.1 delegation draft and the Agent Authorization Envelope (AAE) — currently published as IETF Internet-Draft, also referenced in W3C and A2A discussions — converge on close to identical mechanics:

  • signed delegation token (Ed25519) ↔ signed envelope
  • granted_capabilities ⊆ delegator.effective_capabilities ↔ scope narrowing as MUST
  • depth_cap with depth_remaining-- per hop ↔ depth-limited chain
  • chain_root_passport_id propagated unchanged ↔ principal-authority root reference
  • mandatory expires_at, no never_expires ↔ time-bounded validity
  • per-DT optional revocation_endpoint ↔ per-credential revocation surface

Both the AAE envelope and the principal-authority chain are documented in our recent arXiv paper (2605.06738 — "From Specification to Deployment: Empirical Evidence from a W3C VC + DID Trust Infrastructure for Autonomous Agents"), if you want the architectural framing in writing before the IETF draft revs.

The vocabularies differ (passport_id vs. credential subject, capability dotted-namespace vs. JSON-LD), but the architectural shape lines up cleanly. Worth knowing if v1.1 is still open for external input.

One spec question

The OAP-Passport schema has an optional did field, with did:web:api.aport.io:agents:ap_xxx as the example. Is the slot constrained to did:web, or is the JSON-schema deliberately open to other DID methods? Specifically, would a did:moltrust:xxx or any other resolvable DID method be schema-valid?

If yes, that single field is the cleanest interop point — verifiers with broader DID-resolver support get the upstream principal-authority chain; verifiers without it still see a normal OAP-Passport. No code dependency either direction.

Offer

If delegation.md v1.1 work is still open to external review, I can share:

  • AAE Internet-Draft (current rev) for cross-reference
  • A field-mapping table between AAE envelope shape and the delegation.md token shape (built from the analysis above)
  • Notes on where W3C VC alignment is straightforward and where the source-of-truth boundary stays cleanly on the OAP side

Not pitching a merge, not asking for normative changes. Just offering the mapping work in case it saves time before v1.1 lands.

The complementarity holds regardless: OAP enforces at the tool-call boundary, AAE answers the upstream "does this agent have standing to act for this principal." Different layers, same threat surface.

— Lars, MolTrust

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions