ATS export evaluation

evaluation.integration.test.ts renders deterministic synthetic resume data through current ResumeDocument and buildDocx, then measures extraction with installed PDF.js and DOCX XML. metrics.test.ts locks raw distinct-token recall, order, duplicate, and semantic-grouping behavior, including deliberate drop/duplicate regressions. Grouping compares eligible same-group expected occurrences, preserving repeated values' authored field identity. Link metrics normalize expected and extracted targets and report dropped or changed targets; hidden-leak checks scan paragraph text and link-target channels, including hidden URLs.

Current DOCX output intentionally records its known missing tel: target as a measured loss; any additional dropped or changed target fails integration assertions.

extract.test.ts supplies a positive XML numbering fixture covering paragraph order, numId, level, format, marker, and link-target extraction, plus valid archives with optional entries omitted.

Run from repository root:

pnpm --filter @reactive-resume/tooling test

The integration test writes PDF/DOCX fixtures and raw-count reports to tooling/ats-export-evaluation/test-results/ (ignored test output). Results explicitly distinguish local extraction measurements from vendor parser accuracy. Steps 1–2 add no ATS preset; a later product decision can use measured deficiencies from ats-export-report.md.