Skip to content

P8-A3.1 — research-only typed evidence report bridge prototype #149

Description

@aetherxeg-source

Status

OPEN — RESEARCH_ONLY_TYPED_REPORT_BRIDGE_PROTOTYPE

Parent external gate: #145
Accepted research PR: #147
Selected architecture: REPORT_LAYER_TYPED_HEALTH_BRIDGE

Objective

Build the smallest executable first-party prototype that exports typed evidence/health already known inside ExecSurface without changing the public Alpha.5 raw-v2 contract.

The prototype must consume the internal BackendObservation boundary before typed collection-health information is discarded by the legacy public Observation return path.

Why this location

The external Probity integration demonstrated that reconstructing semantics from serialized raw-v2 forces consumers to rebuild information ExecSurface already knows internally.

A first-party bridge should therefore project existing internal evidence outward; it should not infer stronger semantics post hoc from an already lossy public JSON trace.

Prototype output — bounded

Research report v0 may contain only:

  • producer/report schema identity;
  • backend semantic profile / platform / architecture;
  • typed collection health:
    • complete;
    • incomplete_loss;
    • incomplete_limit;
    • incomplete_capability;
    • incomplete_ambiguity;
    • error where representable;
  • warning/ambiguity codes;
  • bounded proposition records for facts already justified by the backend;
  • typed evidence guarantees such as successful-operation result and runtime-fd path correlation;
  • explicit limitations / unsupported capability state.

Hard prohibitions

The prototype must not:

  • expose new file contents, argv/env, stdin, network payloads or arbitrary memory;
  • add before/after file-state roots;
  • add signing/custody/trusted-time claims;
  • treat backend name as authority;
  • invent exact byte counts not retained by evidence;
  • alter raw schema v2;
  • alter baseline/policy/verdict semantics;
  • alter public CLI or GitHub Action;
  • promote Semantics v3 wholesale;
  • create a new public compatibility promise.

Gate chain

R0 — internal-boundary freeze

Prove the prototype reads only existing BackendObservation / raw observation facts and does not expand acquisition.

R1 — minimal deterministic report

Produce deterministic research-only JSON for:

  • healthy successful fd-write effect;
  • incomplete/ambiguous observation;
  • unsupported capability.

R2 — non-inflation attacks

At minimum:

  • incomplete/warning laundering;
  • backend-name inflation;
  • missing capability;
  • path/process substitution;
  • repeated effects;
  • unsupported exact byte-count/state-root claim;
  • source/profile mismatch.

R3 — live native proof

Use the native Linux x86_64 ptrace backend in test scope and demonstrate that the internal typed health is preserved into the report.

R4 — compatibility

Full CI + P9.3 compatibility must remain green.

R5 — decision

Choose exactly one:

  • BRIDGE_PROTOTYPE_ACCEPTED_FOR_PUBLIC_API_DESIGN
  • BRIDGE_PROTOTYPE_ACCEPTED_RESEARCH_ONLY_MORE_EVIDENCE_REQUIRED
  • BRIDGE_PROTOTYPE_KILLED_AUTHORITY_RISK
  • BRIDGE_PROTOTYPE_KILLED_NO_EXTERNAL_VALUE

Governance

This gate may validate a design path only. Public product/API integration requires a later separately authorized issue/PR and must not be inferred from a successful prototype.

Activity

  1. aetherxeg-source commented on Oct 4, 2026

    @aetherxeg-source
    ContributorAuthor

    R0-R5 closeout candidate

    Accepted pre-closeout head: 7798738f9ebd7a239223d83ac4e5d572ef0df0e9

    Accepted evidence:

    • prototype run 37196488594: SUCCESS, 10/10 tests;
    • full CI 37196488568: SUCCESS;
    • P9.3 37196488586: SUCCESS;
    • adversarial regression 37196488610: SUCCESS on Ubuntu 22.04 and 24.04.

    The live native R3 test passes through the real internal ptrace backend and preserves typed health plus successful fd-write semantics without adding acquisition.

    Retained negatives:

    • 37196376963: formatting-only failure before science;
    • 37196421833: 9/10, with the sole failure caused by a test oracle matching the explicit non-claim string exact_transferred_byte_count; corrected to inspect JSON object keys without weakening any semantic condition.

    R5 candidate:
    BRIDGE_PROTOTYPE_ACCEPTED_FOR_PUBLIC_API_DESIGN

    This is design authorization only, not public integration/release approval and not P8 A5 closeout.

  2. aetherxeg-source commented on Oct 4, 2026

    @aetherxeg-source
    ContributorAuthor

    Final closeout

    PR #150 has been merged after all final-head gates passed.

    Merge SHA:
    c711aa8e1f97cd8310b24411c4e3948c28b01b46

    Final accepted runs on the closeout head:

    • typed report prototype 37196611744: SUCCESS
    • full CI 37196611878: SUCCESS
    • P9.3 Compatibility Contract 37196611689: SUCCESS
    • Adversarial Regression 37196611704: SUCCESS
      • Ubuntu 22.04: PASS
      • Ubuntu 24.04: PASS

    Final R5 decision:

    BRIDGE_PROTOTYPE_ACCEPTED_FOR_PUBLIC_API_DESIGN

    This closes the research-only implementation gate. It authorizes only a separate bounded public-API design exercise.

    It does not expose the report publicly, change observe, change the GitHub Action, alter baseline/policy/verdict semantics, authorize a release, establish production adoption, or satisfy P8 A5.

    P8 #114 remains open because qualifying A1/A4 external execution/use evidence is still missing.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions