Every version, every re-test, in public.
When the underlying model changes, we re-run every skill’s tests and post the result here — pass or fail.
| Skill | Current | Last verified | Local scripts | Model review |
|---|---|---|---|---|
| ASK-00Reporting Diagnostic | v1.0.0 | 8 Oct 2026 | Passing | Pending |
| ASK-01KPI Normalizer | v1.0.0 | 8 Oct 2026 | Passing | Pending |
| ASK-02Evidence-Linked Drafter | v1.0.0 | 8 Oct 2026 | Passing | Pending |
| ASK-03Report QA & Brand Check | v1.0.0 | 8 Oct 2026 | Passing | Pending |
Status legend: Passing — every local script test is green for this version. Pending — the independent review on a Claude (web) configuration hasn’t been published yet. Re-testing — a model or platform change was detected; keep using your current version until a result is posted. Action needed — follow the linked guidance or roll back.
Changelog
- 8 Oct 2026ASK-03v1.0.0Release
Report QA & Brand Check — first release
Flags unsupported claims and banned wording, checks tagged figures against their source, and writes a sign-off checklist.
Test record - 8 Oct 2026ASK-02v1.0.0Release
Evidence-Linked Drafter — first release
Drafts the report narrative with a source tag on every figure, plus an evidence index. Uses approved notes only.
Test record - 8 Oct 2026ASK-01v1.0.0Release
KPI Normalizer — first release
Maps platform exports to pack schema v1 and logs gaps, duplicates and unit mismatches.
Test record - 8 Oct 2026ASK-00v1.0.0Release
Reporting Diagnostic — free
Times one report stage by stage and sorts corrections by source. No client data required.
How we test