• v0.1.0 0ffee85a4b

    ktg released this 2026-08-10 02:45:14 +00:00 | 35 commits to main since this release

    Seven data files, one normative finding contract, one normative corpus spec,
    and 83 conformance cases covering the injection lexicon.

    Seven of the eight JSON artefacts were rebuilt from the commons file alone
    and diffed against their source implementation; three of those against the
    source module at a pinned commit. calibration/calibration.json says
    verified: false about itself, and names the checks that were not run rather
    than attaching a caveat to a pass.

    The corpus holds 83 cases, one per injection-lexicon pattern. Both seeding
    runtimes were measured producing identical lexicon finding sets on all 83,
    through their public entry points. The 13 patterns recorded as divergent
    between the two ports are included unmarked: their divergence lives on
    witness inputs no corpus payload contains.

    Not in this tag, and named rather than papered over: spec/decode-pipeline.md
    needs the decode implementation, and the corpus constrains one of the seven
    data files -- the other tables have no ratified cross-runtime finding id.

    Downloads