Skip to content

PLAN — K1 A2A Handoff / Delegation-Route Evidence (2026 Q2)

  • Current status: Next formal trust-compiler wave after P2c; K1-A freeze path and the first bounded adapter implementation slice are now merged on main and publicly released in v3.4.0; no pack or trust-artifact follow-up shipped in this slice.
  • Date: 2026-03-27
  • Owner: Evidence / Product
  • Inputs: DISCOVERY — Next Evidence Wave, Trust Compiler Audit Matrix, RFC-005, ROADMAP, PLAN-G4, PLAN-P2c
  • Scope (this PR): Formalize the next wave choice only. No adapter code, no pack YAML, no engine work, no Trust Basis or Trust Card changes.

1. Why this plan exists

The trust-compiler line on main is now shipped through P2c. The next step is no longer a vague "post-P2c decision point": maintainers have chosen to formalize the preferred candidate from the discovery note as a named wave.

That choice is:

  • K1 — A2A handoff / delegation-route visibility evidence

K1 is a new wave label for bounded evidence work after the G / P trust-compiler line. It does not imply a pack wave or a new trust-claim surface by itself.

K1 is intentionally:

  • not P2d
  • not "more G4"
  • not MCP authorization-discovery

The bottleneck is still evidence shape, not pack productization. So the next wave stays adapter-first, bounded, and visibility-only.

2. Goal (one sentence)

Add a first-class, bounded canonical evidence seam for A2A handoff / delegation-route visibility, starting from shipped adapter-emitted reality and staying strictly below correctness, trust, or outcome claims.

In this plan, "handoff / delegation-route" refers to one bounded evidence surface, not multiple independent seams.

3. Why K1 now

3.1 What P2c did and did not close

P2c productized the discovery/card line that G4-A made visible. It did not establish a typed seam for handoff / delegation-route evidence.

3.2 Why this is the strongest next bounded surface

The discovery note and audit matrix converge on the same gap:

  • the A2A adapter already has discovery/card visibility
  • the next missing surface is agent-to-agent route / handoff visibility
  • that gap is more about what canonical evidence exists than about pack rules

MCP authorization-discovery remains relevant, but the stronger immediate bottleneck in Assay's shipped evidence line is the A2A route / handoff surface.

3.3 Why this is not "another pack first"

No further A2A pack slice should be treated as default until Assay can observe a bounded, first-class route / handoff seam honestly.

4. Product framing

In scope

  • A2A handoff / delegation-route as a bounded evidence topic
  • Adapter-first work in assay-adapter-a2a once freeze answers are ready
  • One small typed seam in canonical emitted evidence
  • Presence / visibility semantics only
  • Explicit drop rules for ambiguous or blob-like upstream hints

Out of scope

  • Any new pack in the K1 wave
  • Any engine bump
  • Any Trust Basis or Trust Card row in the same slice unless a later bounded predicate is separately justified
  • Handoff success, delegation correctness, route authorization, chain completeness, temporal correctness, or provenance trust
  • Reopening G4-A semantics or extending payload.discovery by default
  • Folding in MCP authorization-discovery; that remains the second candidate, not part of K1

5. Hard language contract

K1 may only claim that route / handoff information is visible in bounded evidence.

K1 must not imply:

  • handoff verified
  • delegation succeeded
  • delegation was allowed
  • the chosen route was correct
  • the target agent was trusted
  • the full delegation chain is complete
  • temporal or cryptographic validity

6. Working v1 seam hypothesis

The likely outcome is one small typed subobject or equivalent bounded field set in canonical A2A payloads. Exact field names are not frozen in this PLAN.

This PLAN only freezes the shape discipline:

  • one seam
  • one meaning
  • typed beats blob
  • visibility beats correctness

Illustrative questions the seam may answer later:

  • is a handoff / delegation-route surface visible at all?
  • did the signal come from typed payload, an allowlisted extension path, or a lossy fallback?
  • is there a bounded target / route kind visibility signal?

These are examples of the kind of seam K1 may freeze later, not a field contract. They are illustrative discovery directions, not provisional field commitments.

7. Phase 0 discovery freeze requirements

Before any implementation PR, K1 must produce a freeze-ready discovery record that answers:

  1. Which upstream A2A flows or fields actually carry route / handoff information?
  2. Which are typed and stable enough for canonical promotion?
  3. Which are only honest as visibility or lossiness signals?
  4. What is the smallest honest seam?
  5. Which upstream hints must not become typed route evidence?

Discovery answers must be grounded in shipped adapter-emitted evidence and current code-path reality, not solely in protocol or standards text.

Minimum artifacts for the later freeze:

  • per-field source mapping table
  • precedence rules when multiple inputs exist
  • representative emitted JSON example
  • negative matrix for false positives
  • explicit may / must-not-imply language

That freeze path now exists as K1-A — Phase 1 formal freeze. It records the first executable contract for a bounded typed handoff seam while keeping K1 itself evidence-first and implementation-free at the plan level.

8. Implementation gates (future, not this PR)

Any future K1 implementation slice should hard-fail review if it:

  • turns route visibility into route correctness
  • guesses from generic task metadata without a frozen source rule
  • promotes arbitrary blob content into typed route evidence
  • silently reuses payload.discovery for a distinct route surface without an explicit freeze decision
  • sneaks in a pack, engine bump, or trust-claim expansion in the same wave
  • widens the seam beyond one bounded route / handoff surface

9. Acceptance for K1-A

K1-A should only count as shipped if all of the following hold:

  1. Canonical emitted A2A evidence gains one bounded route / handoff seam.
  2. The seam is documented with exact source paths and bounded meaning.
  3. Tests show the seam is not promoted from loose or ambiguous input.
  4. Product language stays at visible / observed, not valid / trusted.
  5. No pack or broader trust artifact is shipped in the same wave.

Current status on main: these conditions are now met for the first bounded A2A adapter seam shipped in assay-adapter-a2a. K1-A remains visibility-only and still has no downstream pack or broader trust-artifact follow-up in the same wave.

10. What happens after K1

Only after K1 evidence is real should maintainers revisit whether:

  • a future A2A follow-up pack is honest
  • MCP authorization-discovery should become the next evidence-wave
  • any Trust Basis / Trust Card expansion is justified

Current status on main: that revisit has now selected K2 — MCP authorization-discovery evidence as the planned next bounded evidence wave, while still keeping packs downstream of first-class evidence seams.

No downstream pack or trust-surface follow-up should be assumed as part of K1 closure.

K1 is therefore the next formal evidence wave, not the start of an automatic new pack line.

References