The Scope Creep Automated reviews of new AI-related working group submissions

Update review · November 10, 2026

Actor Chain Overview -01, A Copyedit to Match Protocol -03

draft-mw-oauth-actor-chain-overview -01View on Datatracker →

What changed in -01?

Almost nothing. This revision arrived two weeks after -00 and is a copyedit that syncs the overview with Actor Chain -03.

The design described is the same as in the -00 review. It still has two assurance models, three disclosure modes, six profiles, four claims, three evidence coverage modes, and two delivery modes. The document is still ten pages and still Informational.

There is one substantive change. The overview now says all three coverage modes use envelope version 1.

The version cleanup

In -00, the acte example used "v": 1 for current, and the text said disclosed "uses envelope version 2." The last review pointed out that the version number was acting as a feature flag.

Protocol -03 unified every signed artifact on version 1. The overview follows: "current, disclosed, and path use envelope version 1." The coverage field alone now tells the reader which evidence shape to expect.

Other edits

  • Clearer description of the Verified step. C's proof now "requests access to D and identifies B's accepted step as its predecessor." The -00 wording, "requesting access to D from B's accepted step," was hard to parse.
  • Signature limits are stated more carefully. Signatures "authenticate the signed requests; they do not establish adjacency across omitted actors." current coverage "provides direct verification only of C's signed request."
  • Question 4 was reworded. It now asks whether "the trust assumptions and assurance limits for cross-domain mappings" are clearly specified. The old version asked whether the section "accurately describes mappings' assurance."
  • Section 8 was renamed. "Security and Privacy Considerations" is now "Security Considerations." The text is unchanged, so the privacy statements about correlation and timing are still there.
  • References were cleaned up. The protocol reference now points to -03 and lists all five authors. Draft references carry full draft names.

Was the previous feedback addressed?

Fixed:

  • Envelope versions as feature flags. Gone, as described above.
  • The protocol reference listed four authors. It now lists five and matches this draft.
  • Some compressed sentences. "It hashes C's proof" became "It contains hashes of C's proof." "C's signature additionally needs C's proof" became "verifying C's signature additionally requires C's proof."

Not addressed:

  • Six profiles for three use cases. Table 1 still gives one example per disclosure mode, and none needs Declared assurance. The overview still never says when Declared beats plain Token Exchange.
  • The user is still missing. The workflow still starts with orchestrator A already holding a token. sub is still workflow-subject-alias.
  • Actor key binding. Security Considerations still depends on "independently authorized ActorID bindings" and still doesn't say where they come from.
  • Design Evolution and the abstract lineage. Both are still there. Section 10 was only updated to say -03.
  • MCP, A2A, and Kubernetes framing. Unchanged in Sections 2 and 6.
  • The SD-JWT paragraph. Unchanged.
  • The duplicated sentence in Section 2. "The AS checks the proof and authority, then signs an acceptance. C signs the request; the AS signs its acceptance." is still there.
  • Lowercase claim names starting sentences. Still there, and the new text adds one: "Current evidence can hide A/B. current, disclosed, and path use envelope version 1."
  • Two scenarios for Subset. Table 1 and the A, B, C, D example still use different stories.
  • Overlap with the protocol draft. Protocol -03 kept all nine of its informative sections, so the two documents still cover the same ground twice.
  • "Live interoperability remains untested." Same sentence, same place.

Mechanisms added or removed

Nothing was added. Nothing was removed. Envelope version 2 is gone as a number, and disclosed coverage is fully intact.

The removal list from the -00 review stands as written.

Writing quality

Slightly better. A handful of sentences were expanded so they can be read without the protocol draft open. DPoP is expanded on first use. The Unicode arrows became ASCII.

The overall style is unchanged. Section 5 is still the hardest page, with coverage values, version numbers, and privacy failures packed into one paragraph.

Background of the authors

No change. The same five authors are listed: A. Prasad, Ram Krishnan, Diego R. Lopez, Srinivasa Addepalli, and Suresh Krishnan. Affiliations are still missing from the header.

Questions worth raising

The four questions from the -00 review are all still open. Nothing in this revision changes them:

  1. Do these use cases need a common contract beyond RFC 8693? This is the authors' own Question 1. The WG should answer it once for all the delegation-evidence drafts.
  2. Where does the user enter the chain?
  3. What is Declared assurance for?
  4. Is anyone building this? Protocol -03 changed the version on every signed artifact, and the overview followed within days. That is easy to do when nothing is deployed.

There is no reason to re-read this draft if you read -00. Spend the time on Actor Chain -03, or on the competing inline-proof design in Actor Proofs.