Docs / Tsubasa conformance

Open Stage 1 Conformance Corpus

Public .tsubasa examples and verified expected outputs for Tobi Validator Stage 1.

The conformance corpus makes Stage 1 Tobi Validator behavior inspectable without opening the private validator engine or core implementation.

What it is

A public Stage 1 authoring corridor

The corpus is a public Stage 1 conformance and authoring corridor for .tsubasa artifacts. It contains public-safe source examples, manifest records, canonical-output expectations, diagnostics, and verification evidence.

It is designed to make validator behavior inspectable without exposing the private validator implementation.

How to use it

Inspect first, reproduce with authorized access

  • review the public examples
  • inspect the manifest and verification report
  • use authorized Tobi Validator access to reproduce runs
  • use it as a safe authoring corridor for Stage 1 .tsubasa artifacts
  • do not treat pending cases as verified
  • do not derive final expected outputs without real Tobi runs

What it contains

Public corpus contents

  • public .tsubasa examples
  • accept / reject / equivalence / idempotence / determinism case structure
  • corpus/manifest.v0.1.json
  • real Tobi Validator v0.7.0 verification evidence for 11 cases
  • 1 idempotence case left as PENDING_REAL_TOBI_RUN
  • coverage limitation documentation
  • public/private boundary documentation

What it does not contain

Boundaries that remain closed

  • no Tobi compiler or validator source
  • no validator binary
  • no private fixtures
  • no private golden corpus
  • no private release or distribution logic
  • no full language-coverage claim
  • no certification program
  • no proof, truth, or consensus layer

Current v0.1 status

Verified seed, not a broad coverage claim

  • 12 public cases
  • 11 verified with real Tobi Validator v0.7.0
  • 1 idempotence case pending
  • verified on Windows x86_64 in the current evidence record
  • Linux/macOS cross-platform determinism is not yet claimed
  • v0.1 is a verified seed corpus, not full language coverage