docs(conformance): document the v2.3 report schema (refs TailorAU/tailor-app#5567)

[Tailor-AUS]

Lane Results part 1 of 3 for TailorAU/tailor-app#5567: the v2.3 conformance report format existed only as code. This documents it in the runner README so the two Tailor implementations can publish results docs "in the same schema" faithfully.

What's documented

• Report schema section in spec/v2.3/conformance/runner/README.md:
  • Top-level field table — counts, results, runner_disclaimer, http_coverage (emitted by the runner at src/index.ts main()), plus spec_version and vector_set_ref (stamped by the jq provenance step in .github/workflows/conformance.yml, after the gate runs — with the rationale for stamping at the CI seam rather than in the runner).
  • Per-result field table — path, id, kind, outcome (discriminated union: pass / fail + reason / skip + reason), and the mirrored top-level verification_mode (cryptographic vs structural, per the Honesty disclosure).
• Consuming implementations note: same top-level shape, additive keys allowed (e.g. an implementation identifier object), exclusions recorded as outcome.status: "excluded" + reason — and the exact contract of tools/check-conformance-expected-failures.mjs (reads only .results[].id + .results[].outcome.status + .counts, recomputes counts, accepts only pass/fail/skip, rejects skips), so downstream docs can't break the gate.

Verified against reality

The tables were written from the actual emission code, the workflow's jq step, and the real v23-conformance-report artifact from the latest green main run (run 33471758084, vector_set_ref 902b810): top-level keys, per-result key sets, and outcome shapes in the artifact match the code exactly — no deltas found.

Docs-only: nothing but the README changed; no runner code touched.

🤖 Generated with Claude Code

---

Note

Low Risk
README-only documentation with no runtime, CI, or gate behavior changes.

Overview
Adds a Report schema section to the v2.3 conformance runner README so the --json output and published v23-conformance-report artifact are specified in one place (for TailorAU/tailor-app#5567).

The new docs define top-level fields (counts, results, runner_disclaimer, http_coverage, plus CI-stamped spec_version and vector_set_ref), per-result shape (path, id, kind, outcome, optional verification_mode), and the pass / fail / skip outcome union. It explains that provenance is merged in .github/workflows/conformance.yml after the gate, and why that stays out of the runner.

A Consuming implementations subsection spells out how third parties should publish in the same format (additive keys OK, exclusions as excluded for publication only) versus what check-conformance-expected-failures.mjs actually enforces (id, outcome.status, matching counts, no skips).

The usage example now points --json at this section. Docs-only — no runner or workflow code changes.

^{Reviewed by Cursor Bugbot for commit 9ca183a. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_40889bb8-80dc-4157-b88f-14fc74061aa8)