Evaluate
Runs & Gates
Every run pins the exact case list, versions, and environment it resolved — reproducible for engineers, self-explaining to auditors. Gates enforce suite policy on every pull request, and every override is attributed, justified, and time-boxed.
Release gate
HR Policy Advisor — release gatePR #501 blockedView gate →
Dev loop
WFM Assistant — dev loop$ quality run wfm-assistant-dev-loop
manual · CLI + portal
Release gate
Payroll Copilot — release gatePR #172 blockedView gate →
Release gate
Bryte routing — release gateno runs yet
$ quality run bryte-routing-release-gate
manual · CLI + portal
Release gate
Benefits Explainer — release gateno runs yet
$ quality run benefits-explainer-release-gate
manual · CLI + portal
Run history
every run is a sealed evidence packet| Run | Suite | When | Trigger | Cases | Cost | p95 | Duration | Verdict |
|---|---|---|---|---|---|---|---|---|
| run_5522 | WFM Assistant | 2026-07-07 09:41 | manual | 60✓ · 1✕ | $1.12 | 1.8 s | 3m 12s | pass |
| run_5520 | Payroll Copilot | 2026-07-07 08:15 | pull-request | 26✓ · 2✕ | $2.31 | 2.4 s | 6m 48s | fail |
| run_5518 | HR Policy Advisor | 2026-07-06 22:00 | nightly | 34✓ | $4.62 | 2.7 s | 12m 37s | pass |
| run_5528 | WFM Assistant | 2026-07-07 10:24 | pull-request | 24✓ | $1.34 | 1.9 s | 4m 02s | pass |
| run_5510 | WFM Assistant | 2026-07-06 17:03 | manual | 61✓ | $1.09 | 1.7 s | 2m 54s | pass |
| run_5493 | HR Policy Advisor | 2026-07-06 14:12 | pull-request | 32✓ · 2✕ | $4.87 | 2.9 s | 7m 21s | fail |
| run_5488 | Payroll Copilot | 2026-07-06 11:50 | pull-request | 28✓ | $2.28 | 2.3 s | 6m 05s | pass |
| run_5461 | HR Policy Advisor | 2026-07-05 22:00 | nightly | 34✓ | $4.58 | 2.6 s | 13m 02s | pass |
Each run pins agent build, suite policy, routing table, grader packs, judges, and dataset composition — click through for the sealed packet.
Gate ledger
every decision the gate made — including the overrides2026-07-07 08:15blockedrun_5520 →
quality-gate failed on UKGEPIC/payroll-agents #172 — 0.9.4-rc2 held out of main (functional + adversarial below policy)
2026-07-06 22:00passedrun_5518 →
nightly — 34 / 34 clean; the packet feeds the EU AI Act Art. 15 accuracy evidence
2026-07-06 14:12blockedrun_5493 →
quality-gate failed on UKGEPIC/people-agents #501 — 1.8.2-rc1 blocked at the merge
2026-07-06 11:50mergedrun_5488 →
quality-gate passed on UKGEPIC/payroll-agents #168 — 0.9.3 merged to main
2026-06-30 16:20overridden
Break-glass override — Payroll Copilot 0.9.2 pilot demo build (governance ledger EVT-4790)
requested by bill.w · approved by minerva.m (AI Governance Officer)
Adversarial threshold waived 0.95 → 0.90 for the sandbox pilot only — no production tenants in scope; justification on file.
expires 2026-07-03 (72 h)expired on schedule ✓
This is the same ledger Governance reads — a gate that can be silently overridden is not a gate (§16).
Same suite, three surfaces — the portal Run button, quality run <suite> in any terminal, and the quality-gate GitHub Action run byte-identical policy. Local and CI cannot diverge.