Status
OPEN — P8_A3_EXTERNAL_INTEROP_GAP_REPRODUCTION_GATE
Parent external-validation program: #114
Protocol: docs/development/P8_A3_INTEROPERABILITY_REVIEW_PROTOCOL.md
External source: probityai/agent-evidence-observer#41 (merged 2026-10-02)
External integration path: interop/execsurface-state-2026-10-02/
Public ExecSurface target used externally: v0.1.0-alpha.5
Why this gate exists
An external project integrated the published Alpha.5 native ptrace binary and built a substantial companion layer around the retained raw trace.
The external integration makes concrete technical claims that qualify for P8-A3 processing:
- raw observation schema v2 carries global
complete / warnings but no typed collection-health envelope;
file_descriptor_access write evidence carries operation/fd/path but not the successful byte count or syscall result;
- state before/after bytes and state-root binding therefore had to be captured separately;
- the integration preserves the original ExecSurface trace and explicitly refuses to inflate producer PASS into authority, complete custody, or production adoption;
- the external project reports the resulting collection health as
unknown-no-typed-envelope.
This is a concrete interoperability/evidence-expressivity finding, not generic praise.
Important internal observation
ExecSurface already contains a research-only Semantics v3 prototype with proposition-scoped guarantees and typed completeness dimensions, including observer-health semantics.
That prototype is deliberately disconnected from the public Alpha.5 learn/check path.
The external finding therefore does not authorize immediate schema promotion. It provides external evidence that the research direction addresses a real integration problem.
Gate question
What is the minimum semantics-preserving change, if any, required to let an external evidence consumer determine observation health and successful file-write semantics without rebuilding a bespoke trust layer around raw schema v2?
Required reproduction
G1 — Freeze and reproduce the exact Alpha.5 limitation
Using the immutable published Alpha.5 binary/source:
- reproduce one successful file write;
- prove the retained raw v2 event lacks transferred-byte/result information;
- prove current global
complete + warnings cannot express the typed health state selected by the external consumer;
- retain the raw trace unchanged.
G2 — Map the finding to current Semantics v3
For each external gap, classify it as:
- already represented exactly by the research v3 model;
- represented only partially;
- not represented;
- intentionally out of ExecSurface scope.
Do not change v3 merely to match the external implementation.
G3 — Failure-first adversarial matrix
At minimum:
- zero-byte write / failed write;
- short write;
- repeated writes to the same fd/path;
- observer warning with successful target exit;
- incomplete/truncated observation;
- path/process mismatch;
- shared-fd ambiguity;
- binary/source identity mismatch;
- external before/after snapshot substitution.
G4 — Scope decision
Choose exactly one evidence-backed outcome:
NO_PRODUCT_CHANGE_EXTERNAL_ADAPTER_SUFFICIENT
REPORT_LAYER_TYPED_HEALTH_BRIDGE
RAW_SCHEMA_NEXT_CANDIDATE
SEMANTICS_V3_PUBLIC_PROMOTION_CANDIDATE
INCONCLUSIVE_MORE_EXTERNAL_EVIDENCE_REQUIRED
G5 — Compatibility / anti-drift review
Any candidate change must preserve the P9.3 compatibility contract and must not silently reinterpret Alpha.5 baseline-v2 or verdict semantics.
Non-goals
This gate does not authorize:
- capturing file contents by default;
- building a general filesystem state/Merkle system into ExecSurface;
- turning ExecSurface into a general attestation/custody product;
- claiming external production adoption;
- claiming independent authority from same-operator signatures/timestamps;
- replacing ReproCert or other provenance/attestation layers;
- changing public v1 commitments before an executable proof gate passes.
Current decision
EXTERNAL_GAP_ACCEPTED_FOR_REPRODUCTION_ONLY
No product/schema change is authorized until the exact external finding is reproduced and the bounded decision above is closed.
Status
OPEN —
P8_A3_EXTERNAL_INTEROP_GAP_REPRODUCTION_GATEParent external-validation program: #114
Protocol:
docs/development/P8_A3_INTEROPERABILITY_REVIEW_PROTOCOL.mdExternal source:
probityai/agent-evidence-observer#41(merged 2026-10-02)External integration path:
interop/execsurface-state-2026-10-02/Public ExecSurface target used externally:
v0.1.0-alpha.5Why this gate exists
An external project integrated the published Alpha.5 native ptrace binary and built a substantial companion layer around the retained raw trace.
The external integration makes concrete technical claims that qualify for P8-A3 processing:
complete/ warnings but no typed collection-health envelope;file_descriptor_accesswrite evidence carries operation/fd/path but not the successful byte count or syscall result;unknown-no-typed-envelope.This is a concrete interoperability/evidence-expressivity finding, not generic praise.
Important internal observation
ExecSurface already contains a research-only Semantics v3 prototype with proposition-scoped guarantees and typed completeness dimensions, including observer-health semantics.
That prototype is deliberately disconnected from the public Alpha.5 learn/check path.
The external finding therefore does not authorize immediate schema promotion. It provides external evidence that the research direction addresses a real integration problem.
Gate question
What is the minimum semantics-preserving change, if any, required to let an external evidence consumer determine observation health and successful file-write semantics without rebuilding a bespoke trust layer around raw schema v2?
Required reproduction
G1 — Freeze and reproduce the exact Alpha.5 limitation
Using the immutable published Alpha.5 binary/source:
complete+ warnings cannot express the typed health state selected by the external consumer;G2 — Map the finding to current Semantics v3
For each external gap, classify it as:
Do not change v3 merely to match the external implementation.
G3 — Failure-first adversarial matrix
At minimum:
G4 — Scope decision
Choose exactly one evidence-backed outcome:
NO_PRODUCT_CHANGE_EXTERNAL_ADAPTER_SUFFICIENTREPORT_LAYER_TYPED_HEALTH_BRIDGERAW_SCHEMA_NEXT_CANDIDATESEMANTICS_V3_PUBLIC_PROMOTION_CANDIDATEINCONCLUSIVE_MORE_EXTERNAL_EVIDENCE_REQUIREDG5 — Compatibility / anti-drift review
Any candidate change must preserve the P9.3 compatibility contract and must not silently reinterpret Alpha.5 baseline-v2 or verdict semantics.
Non-goals
This gate does not authorize:
Current decision
EXTERNAL_GAP_ACCEPTED_FOR_REPRODUCTION_ONLYNo product/schema change is authorized until the exact external finding is reproduced and the bounded decision above is closed.