Add bounded cross-platform comparison and interoperability planning

[TimeLordRaps]

Coordinate

Acronyms: application programming interface (API); continuous integration (CI); JavaScript Object Notation (JSON); Verifier Standard (VSTD).

• VSTD numbered profile or closure coordinate: additive VSTD-1 generic-run diagnostic; experimental VSTD-2 modeled-surface analysis and planning side path; Graph candidate/evidence-bound structural precondition.
• Repository release or target commit: v1.3.0 development candidate at 5c42c4c139553b5fa3f2463289c94112c4b76612, stacked on pull request Post-release maintenance: dependencies, onboarding, and examples #30 repair commit 56ea9ab1de4ea34a0dc6beabc753444cd2d2c8fb.
• Claim, schema, or implementation seam: native operating-system result comparison; immutable interoperability component catalog; modeled verification-surface hole detection; registry-bound nonexecuting validation plans.

Change and evidence

• What changes:
  • Reject malformed provenance Graph structure before candidate or evidence-bound Graph computation.
  • Add a supported compare_platform_run_receipts API and vstd compare-platforms command with distinct PASS, CONFLICTED, NOT_ESTABLISHED, and INVALID results.
  • Add native Linux, Windows, and Intel macOS CI observations for the generic-run specimen, plus macOS release-artifact byte comparison.
  • Add an explicitly experimental verifier.interoperability facade for component descriptors, registries, typed modeled-hole analysis, and deterministic plan-only candidate matching.
  • Add a harmless sorted-grocery-list example that exposes five holes, finds two exact candidates, leaves three self-closure holes unmatched, and invokes no checker.
• What does not change: no numbered-profile specification, schema identifier, receipt kind, top-level interoperability export, automatic checker execution, closure result, critical-domain readiness claim, version metadata, tag, or release.
• Falsification condition: reject the comparator claim if a declared platform is absent/duplicated, canonical integrity fails, non-platform bindings differ, or declared results disagree. Reject the planning claim if matching succeeds without exact schema/interaction/relation/mechanism coordinates, if a plan is not bound to the geometry and registry digests, or if planning invokes a checker or claims validation/closure.
• Tests added or updated: exact clean head passed 634 passed, 38 skipped; focused graph 37 passed; comparator/public/presentation 120 passed in an isolated clean commit worktree; interoperability 28 passed; presentation, terminology, acronym, generated-reference, and experiment-index gates passed. Isolated wheel and source distribution build succeeded with pinned build dependencies; both passed Twine metadata checks; a clean wheel exposed the comparator command and 17-name experimental facade.

Consequences

• Serialized-format and compatibility impact: no new receipt schema or wire identifier. The generic example uses the existing open refutation surface to declare platform comparison coordinates. Generic receipt JSON parsing now rejects duplicate keys, non-finite values, and finite-float overflow. Three comparator symbols are additive supported top-level API exports; the planning facade remains experimental and outside verifier.__all__.
• Trust roots, unknowns, residuals, and horizons: local Windows tests do not establish native Linux/Windows/macOS agreement; hosted CI is the next evidence gate. A platform declaration does not prove compatibility or native execution. Planning does not establish availability, checker correctness, evidence, validation, closure, safety, authority, or critical-domain readiness. Strict full-parity VSTD-2 wire loading, execution authorization, evidence acquisition, and post-execution reanalysis remain future gates.
• Downstream documents, schemas, examples, and receipts reviewed: README maturity and usage, changelog, roadmap, architecture, claims and limits, API stability, release procedure, generated reference, generic-run specimen, benign interoperability example, experiment boundary, CI aggregate, and installed-wheel reachability. Normative specifications and schemas are unchanged.

Stack

• Base is codex/post-1.2-maintenance because pull request Post-release maintenance: dependencies, onboarding, and examples #30 must preserve its newcomer improvements while restoring experiments/ as the non-normative roadmap-steering surface.
• Keep this pull request draft and stacked until pull request Post-release maintenance: dependencies, onboarding, and examples #30 is accepted. After that base merges, retarget this pull request to main and re-evaluate the resulting diff and hosted checks.

Checklist

• I did not turn UNKNOWN or CONFLICTED into a clean result.
• I did not strengthen a claim without stronger evidence.
• I did not include secrets, private data, or proprietary operational material.
• Normative text, machine-readable surfaces, examples, and tests agree.
• README maturity, claims guidance, generated reference, and Pages status still agree.