Small tests. Clear boundaries.

Useful guardrails are not proof of a correct conclusion. Here is the testing scope for version 1.0.0.

What was actually run

65 distinct local unit/regression tests; the same 65 passed again after extracting the buyer ZIPs.
12 extracted CLI stages passed with no agent available on PATH.
One small synthetic live-model fixture per Skill passed final validation. Seven calls total included a CaseCover output-format retry and a SheetAudit narrative-caveat retry.
Actual DOCX and one-page text PDF extraction; planted Excel errors and dependency cycles; tampered quotes, answers, change IDs and source-overwrite protection.

These are development tests, not independent certification, customer outcomes or a production accuracy benchmark. No real customer study or cross-platform installation certification is claimed.

What the checks cannot establish

They do not establish source truth, semantic relevance, application correctness, financial assurance, legal validity or a representative research sample. Human interpretation can still be wrong after all structural checks pass.

Alternatives worth considering

Checked 5 November 2026. These products have substantially different scope; this is not a feature-equivalence or savings claim. None of these links is an affiliate link.

Conveyor: broader questionnaire/trust-center platform; its current pricing page shows custom enterprise pricing, not a public dollar entry price.
TestCollab: managed test planning, traceability and AI workflows; Premium is listed at $29/user/month and Elite at $39/user/month on annual billing, with a two-user estimator.
Delve: full qualitative coding/collaboration software, listed at $50/user/month; education pricing differs.
Deckary: broader Office/presentation tooling; Premium listed at $19/month or $180/year.
Draftable Compare: richer document comparison, OCR and desktop workflows; non-legal plan listed at $129/user/year and legal plan at $261/user/year. Volume terms differ.

Manual approved-library lookup, ordinary test-case spreadsheets, manual interview coding, your spreadsheet application's built-in formula tools and standard text diff tools may be sufficient. Do not buy a pack if a tool you already have solves the problem.