feat: add semantic PRD eval suite

[aryeko]

Summary

Adds the Phase 3 deterministic semantic eval suite for define-product. The suite covers PRD contract completeness, stable AC-ID behavior during revision, insufficient-input blocking behavior, product-altitude boundaries, current/future scope separation, source-grounded uncertainty, and downstream citeability.

Also updates @agentic-workflow-kit/eval-kit to github:agentic-workflow-kit/eval-kit#v0.1.2 so disabled model-assisted commands fail closed in this consumer.

Eval Fixture Behavior

• Replaces placeholder exact-text grading with deterministic semantic checks over local Markdown candidates.
• Adds good and bad candidate fixtures for each case.
• Extends pnpm eval:validate-fixtures to fail closed: it validates the expected-check schema, proves malformed check probes throw, requires good fixtures to grade green, and requires bad fixtures to grade red.
• Keeps Promptfoo/model-assisted generation and judging disabled/manual; no model-assisted smokes were run.

Verification

• pnpm install --frozen-lockfile
• pnpm eval:doctor
• pnpm eval:list
• pnpm eval:validate-fixtures
• pnpm check
• git diff --check
• Good deterministic smoke: case-contract-grounded-prd-v1 with candidate-good.md exited 0, verdict green.
• Bad deterministic smoke: case-contract-grounded-prd-v1 with candidate-bad.md exited 1, verdict red.
• Malformed expected-check repro now throws before grading: missing text_includes expectations and missing ac_ids_present.ids both fail closed.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: e8b2f76fc8

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".