Fix #273: align prose + manifest to schema/model 'evaluation' naming

[ealt]

Summary

• Why: The variant evaluation-payload field is evaluation in variant.schema.json, the eden_contracts.Variant model, and the storage/wire impl, but the spec prose spelled it metrics across eight chapters plus the integrator-manifest table. Two names for one field is a readability/onboarding hazard and a latent parity trap — the schema-parity job only checks schema↔model (both already evaluation), so the prose metrics naming was unguarded.
• Option 1 (selected during triage): align prose + manifest to the schema/model evaluation naming, keeping the wire + on-tree manifest key stable. The reference integrator already emitted evaluation in .eden/variants/<id>/evaluation.json (see _manifest.py), so this is a prose-and-docstring correction, not a manifest-shape change.
• Scope was wider than the issue's two named locations. The issue named 02-data-model.md §9.1 and 06-integrator.md §4.2, but a full audit (per the AGENTS.md "spec inter-chapter restatement is a conflict surface" pitfall) found the same field-name spelling restated across eight chapters (02/03/04/05/06/07/08/10). Renaming only the two named spots would have relocated the drift, so every backtick'd field reference is renamed in lockstep.
• Latent conformance bug fixed: test_evaluator_submission.py asserted variant.get("metrics") is None in two evaluation_error scenarios — always None on the wire (field is evaluation), so those assertions passed trivially and never tested the no-graft guarantee they claimed. Corrected to variant.get("evaluation").

What this does NOT cover

• baseline.metrics config block (02-data-model.md §2.7, experiment-config.schema.json, eden_contracts.config.BaselineConfig.metrics) — deliberately keeps the metrics name. It is a distinct config field that writes into variant.evaluation; renaming it would touch the config surface, which Option 1 explicitly keeps stable. Not a deferral — an intentional boundary.
• Plain-English "metric" concept uses (metric values §1.3, metric names in the evaluation schema §8, "objective over metrics") — left untouched. A metric is a real domain concept; the field that holds the metrics is evaluation.
• docs/conformance-coverage.md — a non-CI-enforced generated snapshot, last regenerated at Issue #99: Per-claim MUST/SHOULD audit (chapters 02/03/04/05/06/07/08/09) #112 and drifted ~40 keyword lines since Evaluatable baseline variant (seed becomes a kind=baseline Variant) #122. Regenerating it in this PR would dump unrelated Evaluatable baseline variant (seed becomes a kind=baseline Variant) #122-era churn into a focused rename, so it is left for routine regeneration (it will pick up the renamed prose then). No tracked deferral — the doc self-documents that it is generated and re-runnable.
• No new runtime behavior: the only impl changes are docstrings + one human-readable validate_acceptance reason string. No wire/schema/model field changes.

Fresh-operator walkthrough

N/A — internal change only (spec prose, docstrings, one error-message string, test data). No operator-facing surface changes. The 07-wire-protocol.md §11 reference-helper endpoint path was corrected to /validate/evaluation to match the impl, which already exposed that path — no behavior change.

Test plan

All literal pre-push gates from AGENTS.md "Commands" run locally:

• python3 scripts/check-rename-discipline.py — clean
• uv run ruff check . — clean
• uv run pyright — 0 errors
• markdownlint-cli2 (spec + CHANGELOG + roadmap) — 0 errors
• check-jsonschema --check-metaschema — ok
• uv run pytest reference/packages/eden-contracts/tests/test_schema_parity.py — 219 passed
• python3 scripts/spec-xref-check.py — all 1105 §-refs resolve
• check_citations.py — 267 scenarios cite valid spec MUSTs
• uv run pytest -q — 2121 passed, 230 skipped, 3 failed (all pre-existing/environmental: error: 1Password: failed to fill whole buffer on git commit in temp-repo tests — local SSH-signing lock, unrelated to this change; none in touched files)
• uv run pytest -q conformance/ -n auto — 255 passed, 13 skipped
• bash reference/compose/healthcheck/smoke.sh — PASS (includes "variant/* refs published to forgejo", exercising the integrator evaluation manifest end-to-end)

Related issues

• Closes Spec/impl drift: variant evaluation-payload field is evaluation (schema/model) vs metrics (02-data-model §9.1 prose + integrator manifest) #273 — Spec/impl drift: variant evaluation-payload field is evaluation (schema/model) vs metrics (prose + manifest)
• Refs Evaluatable baseline variant (seed becomes a kind=baseline Variant) #122 — surfaced this drift (its CHANGELOG entry's "Pre-existing drift surfaced (tracked)" note)

🤖 Generated with Claude Code