Skip to content

P5_CLOSEOUT_DECISION.md

Latest commit

 

History

History
100 lines (70 loc) · 4.54 KB

P5_CLOSEOUT_DECISION.md

File metadata and controls

100 lines (70 loc) · 4.54 KB
Raw

ExecSurface — P5 Runtime Attestation & Provenance Closeout Decision

Date: 2026-09-30 Parent program: #100 Protocol: docs/development/P5_CLOSEOUT_PROTOCOL.md Branch: development/post-alpha4-behavioral-integrity

Decision

P5_EXISTING_STANDARDS_COMPOSITION_SUFFICIENT_BOUNDED

Retained A5 result:

P5_A5_CROSS_ATTESTATION_GRAPH_PASS_BOUNDED

Scientific wording boundary:

No demonstrated need for a new custom predicate/schema in the bounded P5 experiment.

This is not a universal representation claim and does not state that custom domain data is absent. ExecSurface-specific SCAI attribute/condition data remains domain-specific while using existing standard predicate structures.

Independent closeout evidence

Successful closeout workflow:

  • source SHA: c3caa857581d6370008fe30779bd5fd084b28751
  • workflow: .github/workflows/p5-closeout.yml
  • run: 36760093433
  • conclusion: success
  • evidence artifact ID: 11118009105
  • artifact name: p5-closeout-36760093433-1
  • artifact digest: sha256:94312e440d6873bb3577ada92e3e0f13b52d103ccf63db962c8d367a8c52a815
  • artifact size: 38377 bytes
  • artifact expiry at capture: 2026-10-30T18:39:18Z

The closeout run independently verified:

  1. predecessor ancestry from P4 through P5-A4 and both A5 evidence sources;
  2. immutable public v0.1.0-alpha.4 and v0.1 source 48e0b9a0707553349e75e97a9dfa096d13f9ab5d;
  3. no public crates/ changes introduced by P5 research after P4 closeout;
  4. exact successful A5 run and artifact metadata/digest;
  5. exact retained historical A5 counterexample run and artifact metadata/digest;
  6. internal evidence-file checksums for both A5 artifacts;
  7. the frozen 12/12 successful A5 result and candidate marker;
  8. the retained 8/12 historical counterexample result;
  9. mandatory ExecSurface attestation payload representation;
  10. existing Runtime Trace, SCAI, SVR, and SLSA predicate types in the research model;
  11. independent live re-verification of the accepted P5-A4 GitHub/Sigstore attestation.

Retained negative evidence

The first scientifically executable P5-A5 run remains part of the scientific record:

  • source SHA: a7753bee6cc9f145adc7823dd82837fa40cfca01
  • run: 36755962689
  • frozen result: 8 passed / 4 failed
  • artifact ID: 11116268894
  • artifact digest: sha256:97aa4db53e27355d8b2d62e86d6f1ba5c2ee6d7d6e10607b965e4730fa878314

The four surviving attacks were baseline substitution, current-surface substitution, source substitution, and duplicate semantic SVR property. They were classified as implementation-level cross-attestation binding/canonicalization gaps, not evidence of a standards representation gap. The exact 12-test corpus and acceptance threshold were not weakened.

Corrected A5 evidence

Successful corrected A5 source/run:

  • source SHA: d6d0d45b686ed8c9332ea6b0924537a345049bfb
  • run: 36758531046
  • result: 12/12 PASS
  • artifact ID: 11117568130
  • artifact digest: sha256:d1ba8f88a42b0b3f23880445f6063735cef99b306861241ed2e04cdeee57d343

The same run also re-proved accepted P5-A4/A3/A2/A1/A0, critical P4 authority gates, Semantics v3, and public M11.

What P5 establishes, bounded

For the tested research scope, ExecSurface can represent and bind its runtime-verification result using a composition of:

  • in-toto Statement v1;
  • Runtime Trace v0.1;
  • SCAI v0.3;
  • SVR v0.2;
  • optional exact SLSA provenance v1 reference;
  • GitHub/Sigstore artifact-attestation identity and verification.

The composition preserves explicit subject, source/artifact/workflow context, baseline/current-surface digests, observer/capability/completeness, policy, verdict, evidence, provenance, and verifier identity under the tested graph verifier.

Cryptographic/signature validity remains separate from semantic authority. Signatures, provenance, backend names, and standards labels do not upgrade an otherwise insufficient behavioral-verification result.

Non-claims / promotion boundary

This decision does NOT authorize:

  • public integration or release promotion;
  • changes to main, alpha.4 semantics, stable @v0.1, public v2 meaning, or the default observer;
  • a universal claim that the standards composition represents every runtime-verification system;
  • a claim that generic consumers automatically understand ExecSurface proposition authority;
  • a SLSA level claim;
  • deletion or reinterpretation of negative evidence.

P5 is therefore scientifically closed, bounded, research-only. The next parent-program phase is P6 — Competitive Falsification, which must compare overlapping factual propositions rather than produce a marketing winner score.