Skip to content

research: add guarded GSAE-E0 Stage-A conformance apparatus - #7

Open
ndrorchestration wants to merge 21 commits into
design/cross-runtime-contract-v1from
research/gsae-e0-stage-a-harness
Open

ndrorchestration wants to merge 21 commits into
design/cross-runtime-contract-v1from
research/gsae-e0-stage-a-harness

Conversation

@ndrorchestration

@ndrorchestration ndrorchestration commented Sep 16, 2026

Copy link
Copy Markdown
Owner

Summary

Implements the bounded Stage-A conformance apparatus tracked in #6 on top of PR #5's exact contract branch.

This is a stacked research-apparatus PR. Base is design/cross-runtime-contract-v1; source under test remains PR #5 exact head 07a09698ca66e8837d04e6ec05b4de3448eced04.

Added

  • isolated experiments/gsae_e0/ research package;
  • 28-fixture frozen Stage-A candidate and deterministic loader;
  • typed fixture/result/evidence schemas;
  • structured-semantic classification for AUTH-01AUTH-10;
  • ACP-native behavior and fail-closed negative probes;
  • exact source ancestry + ACP-source-drift verification;
  • deterministic result/evidence identities and retry lineage;
  • guarded in-memory Stage-A orchestration requiring explicit execution identities;
  • synthetic-only runner tests;
  • README with explicit evidence ceiling;
  • pytest discovery for the research tests and full-history checkout needed for source-binding proof.

Verification

Exact apparatus head: 2548ec7a2072d848360481fcf972270204f8aff8.

Push workflow run 35064353443 completed SUCCESS across Python 3.10–3.14.

Fresh PR-event workflow run 35065405012 also completed SUCCESS on the same exact head across Python 3.10–3.14; representative job reports 126 passed, 0 failed.

Comparison from PR #5 source head to the apparatus head shows no changes under src/agent_control_plane/.

Evidence ceiling

This PR can establish only that the Stage-A apparatus is implemented, deterministic under its tests, source-bound, and ready for review.

It does not establish or authorize:

  • a canonical GSAE-E0 Stage-A run;
  • contract feasibility or portability;
  • DGAF/PDMAL efficacy or authorization;
  • production/security certification;
  • empirical/scientific N > 0.

Current research state: R2 FROZEN / APPARATUS_READY / EXECUTION NOT AUTHORIZED / NOT RUN / N=0.

Issue #6 is closed as the completed apparatus-build record. PR #7 remains the stacked review surface. R2 freeze acceptance is separate from execution authorization.

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

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant