Skip to content

Trust Compiler Audit Matrix (2026-03-26)

This matrix records the trust-compiler line that moved from T1b through K2-A Phase 1, plus the discovery-only ranking that led first to K1 and then to K2. It is meant as an audit aid: each row captures the wave, the PRs that carried it, the kernel files, the contract/result that matters, and whether that row is shipped on main and/or part of the public v3.3.0 / v3.4.0 / v3.5.0 release truth.

Wave PR(s) Kernbestanden Belangrijk contract / resultaat main shipped? Release-truth?
T1b #923 crates/assay-evidence/src/trust_card.rs, crates/assay-cli/src/cli/commands/trustcard.rs Deterministic trustcard.json / trustcard.md derived from trust-basis.json; geen tweede semantische classificatielaag Yes Yes — in v3.3.0
G3 #929 crates/assay-evidence/src/g3_authorization_context.rs, crates/assay-evidence/src/trust_basis.rs, crates/assay-evidence/src/trust_card.rs authorization_context_visible; Trust Basis met 7 claims; Trust Card schema_version: 2; bounded auth_scheme / auth_issuer / principal op ondersteunde MCP decision evidence Yes Yes — in v3.3.0
Outward positioning #930 README.md, trust-compiler docs slices README / outward positioning verschuift naar claims-as-code / trust compiler taal Yes Yes — public docs line through v3.4.0
P2a + engine 1.2 #935 crates/assay-evidence/src/lint/packs/checks.rs, crates/assay-evidence/src/lint/packs/mod.rs, packs/open/mcp-signal-followup/pack.yaml Built-in mcp-signal-followup; g3_authorization_context_present; pack engine 1.2 Yes Yes — in v3.3.0
H1 #937 docs/architecture/MIGRATION-TRUST-COMPILER-3.2.md, crates/assay-evidence/tests/h1_trust_kernel_alignment.rs Migration SSOT; kernel / Trust Basis / Trust Card alignment hardening Yes Yes — in v3.3.0
H1a #938 shared trust-kernel test vectors and companion tests Shared vectors and alignment support without new product semantics Yes Yes — in v3.3.0
Roadmap sync #939 docs/ROADMAP.md Roadmap sync after P2a / H1 / P2b ordering hardens Yes Yes — public docs line through v3.4.0
P2b #940 packs/open/a2a-signal-followup/pack.yaml, built-in pack wiring Built-in a2a-signal-followup (A2A-001..003); presence-only on canonical A2A adapter evidence Yes Yes — in v3.3.0
v3.3.0 line release cut + #941 hardening CHANGELOG.md, docs/architecture/RELEASE-PLAN-TRUST-COMPILER-3.3.md, crates.io line First coherent public trust-compiler line: Trust Basis, Trust Card v2 / 7 claims, G3, engine 1.2, built-in P2a + P2b, migration SSOT Yes Yes — tag / GitHub release / crates.io
G4 planning #942, #943 docs/architecture/PLAN-G4-A2A-DISCOVERY-CARD-EVIDENCE-2026q2.md, docs/architecture/G4-A-PHASE1-FREEZE.md Discovery/card evidence wave framed as adapter-first evidence seam, not pack-first expansion Yes Yes — public docs line through v3.4.0
G4-A Phase 1 #944, #945 crates/assay-adapter-a2a/src/adapter_impl/discovery.rs, A2A payload wiring New typed payload.discovery seam for A2A discovery/card visibility Yes Yes — in v3.4.0
P2c planning + freeze #946, #947, #948 docs/architecture/PLAN-P2c-A2A-DISCOVERY-CARD-FOLLOWUP-PACK.md and related status docs P2c frozen as a follow-on pack to G4-A, not part of G4 itself Yes Yes — public docs line through v3.4.0
P2c #949 built-in a2a-discovery-card-followup, pack checks, open mirror Built-in A2A-DC-001 / A2A-DC-002; json_path_exists + value_equals; requires.assay_min_version: ">=3.3.0" Yes Yes — in v3.4.0
K1 formalization #989, #990, #991 docs/architecture/PLAN-K1-A2A-HANDOFF-DELEGATION-ROUTE-EVIDENCE-2026q2.md, docs/ROADMAP.md, docs/architecture/RFC-005-trust-compiler-mvp-2026q2.md Next formal wave chosen as bounded A2A handoff / delegation-route visibility evidence; explicitly adapter-first, evidence-first, no pack in the same wave Yes Yes — public docs line through v3.4.0
K1-A Phase 1 freeze + first adapter slice + closure #992, #994, #995 docs/architecture/K1-A-PHASE1-FREEZE.md, crates/assay-adapter-a2a/src/adapter_impl/handoff.rs, crates/assay-adapter-a2a/src/adapter_impl/payload.rs, crates/assay-adapter-a2a/src/adapter_impl/tests.rs, trust-sync docs Top-level payload.handoff seam always present; typed_payload / unknown only; positive only for task.requested + task.kind == "delegation"; explicit non-promotion for task.updated, artifact.shared, generic-message fallback, synthetic unknown-task, and non-delegation task.kind Yes Yes — in v3.4.0
K2 planning + K2-A freeze docs-only follow-up on main docs/architecture/PLAN-K2-MCP-AUTHORIZATION-DISCOVERY-EVIDENCE-2026q2.md, docs/architecture/K2-A-PHASE1-FREEZE.md, docs/architecture/K2-A-PHASE1-FREEZE-PREP.md, docs/ROADMAP.md, docs/architecture/RFC-005-trust-compiler-mvp-2026q2.md K2 is formalized as the next bounded MCP authorization-discovery wave; K2-A freezes one seam, repo-reality-first source classes, and hard non-promotion rules before implementation Yes Yes — public docs line through v3.5.0
K2-A Phase 1 implementation #1004 crates/assay-core/src/mcp/types.rs, crates/assay-core/src/mcp/parser.rs, crates/assay-core/src/mcp/mapper_v2.rs, crates/assay-core/tests/mcp_transport_compat.rs, crates/assay-cli/tests/mcp_transport_import.rs Bounded MCP authorization-discovery seam on imported MCP traces: top-level episode_start.meta.mcp.authorization_discovery, visibility-only, positive only for typed WWW-Authenticate discovery on supported 401 transport paths; explicit non-promotion without typed runtime-observed provenance Yes Yes — in v3.5.0
Discovery-only next-wave note discovery note on main docs/architecture/DISCOVERY-NEXT-EVIDENCE-WAVE-2026Q2.md Preferred next evidence-wave candidate: A2A handoff / delegation-route visibility; second candidate: MCP authorization-discovery; explicitly not automatically another pack Yes Yes — historical public docs context through v3.5.0

Current call

The trust-compiler line is now public through K2-A Phase 1 in v3.5.0. The current documentation converges on this call:

  • keep G4-A limited to post-merge verification / release-truth hygiene
  • treat Assay as a CI-native protocol-governance and evidence-first product line, not a trust-score-first engine
  • current public slice now includes bounded MCP authorization-discovery visibility evidence via K2-A
  • active bounded evidence wave: K2 — MCP authorization-discovery evidence
  • K2-A Phase 1 is now public in v3.5.0
  • v3.5.0 release status does not imply an automatic second K2 slice
  • keep protocol packs downstream of first-class evidence seams, not as the default next move

Post-release truth-check (v3.5.0)

Checked and aligned for K2-A public release truth:

  • CHANGELOG.md now treats K2-A as publicly shipped in v3.5.0
  • README.md now lists K2-A inside the current public trust-compiler line
  • docs/ROADMAP.md now reflects K2-A as public in v3.5.0, not main-only
  • docs/architecture/RFC-005-trust-compiler-mvp-2026q2.md now treats K2-A as part of the active public line
  • docs/architecture/PLAN-K2-MCP-AUTHORIZATION-DISCOVERY-EVIDENCE-2026q2.md now describes K2-A with public release status, not main-only status
  • docs/architecture/K2-A-PHASE1-FREEZE.md now treats the Phase 1 seam as a released v3.5.0 slice
  • docs/architecture/index.md now points at K2-A as public in v3.5.0

See also: