spec-evaluate — Acceptance Test Runner

Execute the acceptance test plan (test.md) against the built feature, capture evidence, and report requirement-level pass/fail as findings that feed back into the implementation loop.

Language Rules

1. Auto-detect input language → output in the same language
2. Japanese input → Japanese output, use the *.ja.md reference files
3. English input → English output, use the *.md reference files
4. Explicit override takes priority

The evaluator instruction sheet is references/evaluator-prompt.md (.ja.md for Japanese). The result file format is references/result-format.md (.ja.md for Japanese).

Options